# Zeus Strategy: First Generation of GodStra Strategy with maximum # AVG/MID profit in USDT # Author: @Mablue (Masoud Azizi) # github: https://github.com/mablue/ # IMPORTANT: INSTALL TA BEFOUR RUN(pip install ta) # freqtrade hyperopt --hyperopt-loss SharpeHyperOptLoss --spaces buy sell roi --strategy Zeus # --- Do not remove these libs --- from datetime import timedelta, datetime from freqtrade.persistence import Trade from freqtrade.strategy import (BooleanParameter, CategoricalParameter, DecimalParameter, stoploss_from_open, IntParameter, IStrategy, merge_informative_pair, informative, stoploss_from_absolute) import pandas as pd import numpy as np import os import json from pandas import DataFrame from typing import Optional, Union, Tuple import math import logging from pathlib import Path # -------------------------------- # Add your lib to import here test git import ta import talib.abstract as talib import freqtrade.vendor.qtpylib.indicators as qtpylib from datetime import timezone, timedelta logger = logging.getLogger(__name__) # Machine Learning from sklearn.model_selection import train_test_split import joblib import matplotlib.pyplot as plt from sklearn.metrics import ( classification_report, confusion_matrix, accuracy_score, roc_auc_score, roc_curve, precision_score, recall_score, precision_recall_curve, f1_score, mean_squared_error, r2_score ) from sklearn.tree import export_text import inspect from sklearn.feature_selection import SelectFromModel from tabulate import tabulate from sklearn.feature_selection import VarianceThreshold import seaborn as sns import lightgbm as lgb from sklearn.model_selection import cross_val_score import optuna.visualization as vis import optuna from lightgbm import LGBMRegressor from sklearn.metrics import mean_squared_error from sklearn.model_selection import train_test_split from sklearn.linear_model import LinearRegression, Ridge, HuberRegressor from sklearn.preprocessing import StandardScaler, PolynomialFeatures from sklearn.pipeline import make_pipeline from sklearn.svm import SVR from sklearn.ensemble import RandomForestRegressor from sklearn.ensemble import GradientBoostingRegressor from sklearn.preprocessing import StandardScaler from sklearn.ensemble import HistGradientBoostingRegressor from sklearn.impute import SimpleImputer from sklearn.pipeline import Pipeline # Tensorflow import pandas as pd import numpy as np import tensorflow as tf from tensorflow.keras import layers, models from tensorflow.keras.models import load_model from keras.utils import plot_model from keras.models import Sequential from keras.layers import LSTM, Dense from sklearn.preprocessing import MinMaxScaler from tensorflow.keras.models import Sequential from tensorflow.keras.layers import LSTM, Dense from tensorflow.keras.optimizers import Adam os.environ["CUDA_VISIBLE_DEVICES"] = "-1" # désactive complètement le GPU os.environ["TF_XLA_FLAGS"] = "--tf_xla_enable_xla_devices=false" # Couleurs ANSI de base RED = "\033[31m" GREEN = "\033[32m" YELLOW = "\033[33m" BLUE = "\033[34m" MAGENTA = "\033[35m" CYAN = "\033[36m" RESET = "\033[0m" import warnings warnings.filterwarnings( "ignore", message=r".*No further splits with positive gain.*" ) def pprint_df(dframe): print(tabulate(dframe, headers='keys', tablefmt='psql', showindex=False)) def normalize(df): df = (df - df.min()) / (df.max() - df.min()) return df class Zeus_TensorFlow(IStrategy): startup_candle_count = 24 * 12 # Machine Learning model = None model_indicators = [] indicator_target = 'mid_smooth_5' # Tensorflow lookback = 60 future_steps = 12 y_no_scale = False path = f"user_data/plots/" # ROI table: minimal_roi = { "0": 0.564, "567": 0.273, "2814": 0.12, "7675": 0 } # Stoploss: stoploss = -1 # 0.256 # Custom stoploss use_custom_stoploss = False trailing_stop = True trailing_stop_positive = 0.15 trailing_stop_positive_offset = 0.20 trailing_only_offset_is_reached = True # Buy hypers timeframe = '5m' max_open_trades = 5 max_amount = 40 parameters = {} # DCA config position_adjustment_enable = True plot_config = { "main_plot": { "sma24_1h": { "color": "pink" }, "sma5_1d": { "color": "blue" }, # "sma24": { # "color": "yellow" # }, "sma60": { "color": "green" }, "bb_lowerband": { "color": "#da59a6"}, "bb_upperband": { "color": "#da59a6", }, # "sma12": { # "color": "blue" # }, "mid_smooth_3_1h": { "color": "blue" } }, "subplots": { "Rsi": { "max_rsi_24": { "color": "blue" }, "max_rsi_24_1h": { "color": "pink" }, # "rsi_1h": { # "color": "red" # }, # "rsi_1d": { # "color": "blue" # } }, "Rsi_deriv1": { "sma24_deriv1_1h": { "color": "pink" }, "sma24_deriv1": { "color": "yellow" }, "sma5_deriv1_1d": { "color": "blue" }, "sma60_deriv1": { "color": "green" } }, "Rsi_deriv2": { "sma24_deriv2_1h": { "color": "pink" }, "sma24_deriv2": { "color": "yellow" }, "sma5_deriv2_1d": { "color": "blue" }, "sma60_deriv2": { "color": "green" } }, 'Macd': { "macd_rel_1d": { "color": "cyan" }, "macdsignal_rel_1d": { "color": "pink" }, "macdhist_rel_1d": { "color": "yellow" } } } } columns_logged = False pairs = { pair: { "first_buy": 0, "last_buy": 0.0, "last_min": 999999999999999.5, "last_max": 0, "trade_info": {}, "max_touch": 0.0, "last_sell": 0.0, 'count_of_buys': 0, 'current_profit': 0, 'expected_profit': 0, 'previous_profit': 0, "last_candle": {}, "last_count_of_buys": 0, 'base_stake_amount': 0, 'stop_buy': False, 'last_date': 0, 'stop': False, 'max_profit': 0, 'total_amount': 0, 'has_gain': 0, 'force_sell': False, 'force_buy': False } for pair in ["BTC/USDC", "ETH/USDC", "DOGE/USDC", "XRP/USDC", "SOL/USDC", "BTC/USDT", "ETH/USDT", "DOGE/USDT", "XRP/USDT", "SOL/USDT"] } # 20 20 40 60 100 160 260 420 # 50 50 100 300 500 # fibo = [1, 1, 2, 3, 5, 8, 13, 21] # my fibo # 50 50 50 100 100 150 200 250 350 450 600 1050 fibo = [1, 1, 1, 2, 2, 3, 4, 5, 7, 9, 12, 16, 21] baisse = [1, 2, 3, 5, 7, 10, 14, 19, 26, 35, 47, 63, 84] # Ma suite 1 1 1 2 2 3 4 5 7 9 12 16 21 # Mise 50 50 50 100 100 150 200 250 350 450 600 800 1050 # Somme Mises 50 100 150 250 350 500 700 950 1300 1750 2350 3150 4200 # baisse 1 2 3 5 7 10 14 19 26 35 47 63 84 # factors = [1, 1.1, 1.25, 1.5, 2.0, 3] # thresholds = [2, 5, 10, 20, 30, 50] factors = [0.5, 0.75, 1, 1.25, 1.5, 2] thresholds = [0, 2, 5, 10, 30, 45] trades = list() max_profit_pairs = {} mise_factor_buy = DecimalParameter(0.01, 0.1, default=0.05, decimals=2, space='buy', optimize=True, load=True) indicators = {'sma5', 'sma12', 'sma24', 'sma60'} indicators_percent = {'percent', 'percent3', 'percent12', 'percent24', 'percent_1h', 'percent3_1h', 'percent12_1h', 'percent24_1h'} mises = IntParameter(1, 50, default=5, space='buy', optimize=True, load=True) ml_prob_buy = DecimalParameter(-0.5, 0.5, default=0.0, decimals=2, space='buy', optimize=True, load=True) ml_prob_sell = DecimalParameter(-0.5, 0.5, default=0.0, decimals=2, space='sell', optimize=True, load=True) pct = DecimalParameter(0.005, 0.05, default=0.012, decimals=3, space='buy', optimize=True, load=True) pct_inc = DecimalParameter(0.0001, 0.003, default=0.0022, decimals=4, space='buy', optimize=True, load=True) rsi_deb_protect = IntParameter(50, 90, default=70, space='protection', optimize=True, load=True) rsi_end_protect = IntParameter(20, 60, default=55, space='protection', optimize=True, load=True) sma24_deriv1_deb_protect = DecimalParameter(-4, 4, default=-2, decimals=1, space='protection', optimize=True, load=True) sma24_deriv1_end_protect = DecimalParameter(-4, 4, default=0, decimals=1, space='protection', optimize=True, load=True) # ========================================================================= should_enter_trade_count = 0 def confirm_trade_entry(self, pair: str, order_type: str, amount: float, rate: float, time_in_force: str, current_time: datetime, entry_tag: Optional[str], **kwargs) -> bool: minutes = 0 if self.pairs[pair]['last_date'] != 0: minutes = round(int((current_time - self.pairs[pair]['last_date']).total_seconds() / 60)) dataframe, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe) last_candle = dataframe.iloc[-1].squeeze() last_candle_2 = dataframe.iloc[-2].squeeze() last_candle_3 = dataframe.iloc[-3].squeeze() condition = True #(last_candle[f"{indic_5m}_deriv1"] >= indic_deriv1_5m) and (last_candle[f"{indic_5m}_deriv2"] >= indic_deriv2_5m) allow_to_buy = True #(condition and not self.pairs[pair]['stop']) | (entry_tag == 'force_entry') if allow_to_buy: self.trades = list() self.pairs[pair]['first_buy'] = rate self.pairs[pair]['last_buy'] = rate self.pairs[pair]['max_touch'] = last_candle['close'] self.pairs[pair]['last_candle'] = last_candle self.pairs[pair]['count_of_buys'] = 1 self.pairs[pair]['current_profit'] = 0 self.pairs[pair]['last_max'] = max(last_candle['close'], self.pairs[pair]['last_max']) self.pairs[pair]['last_min'] = min(last_candle['close'], self.pairs[pair]['last_min']) dispo = round(self.wallets.get_available_stake_amount()) self.printLineLog() stake_amount = self.adjust_stake_amount(pair, last_candle) self.pairs[pair]['total_amount'] = stake_amount self.log_trade( last_candle=last_candle, date=current_time, action=("🟩Buy" if allow_to_buy else "Canceled") + " " + str(minutes), pair=pair, rate=rate, dispo=dispo, profit=0, trade_type=entry_tag, buys=1, stake=round(stake_amount, 2) ) return allow_to_buy def confirm_trade_exit(self, pair: str, trade: Trade, order_type: str, amount: float, rate: float, time_in_force: str, exit_reason: str, current_time, **kwargs, ) -> bool: # allow_to_sell = (minutes > 30) dataframe, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe) last_candle = dataframe.iloc[-1].squeeze() minutes = int(round((current_time - trade.open_date_utc).seconds / 60, 0)) profit =trade.calc_profit(rate) force = self.pairs[pair]['force_sell'] allow_to_sell = minutes > 30 and (last_candle['hapercent'] < 0 ) or force or (exit_reason == 'force_exit') or (exit_reason == 'stop_loss') if allow_to_sell: self.trades = list() self.pairs[pair]['last_count_of_buys'] = trade.nr_of_successful_entries # self.pairs[pair]['count_of_buys'] self.pairs[pair]['last_sell'] = rate self.pairs[pair]['last_candle'] = last_candle self.pairs[pair]['max_profit'] = 0 self.pairs[pair]['previous_profit'] = 0 self.trades = list() dispo = round(self.wallets.get_available_stake_amount()) # print(f"Sell {pair} {current_time} {exit_reason} dispo={dispo} amount={amount} rate={rate} open_rate={trade.open_rate}") self.log_trade( last_candle=last_candle, date=current_time, action="🟥Sell " + str(minutes), pair=pair, trade_type=exit_reason, rate=last_candle['close'], dispo=dispo, profit=round(profit, 2) ) self.pairs[pair]['force_sell'] = False self.pairs[pair]['has_gain'] = 0 self.pairs[pair]['current_profit'] = 0 self.pairs[pair]['total_amount'] = 0 self.pairs[pair]['count_of_buys'] = 0 self.pairs[pair]['max_touch'] = 0 self.pairs[pair]['last_buy'] = 0 self.pairs[pair]['last_date'] = current_time self.pairs[pair]['current_trade'] = None # else: # self.printLog(f"{current_time} SELL triggered for {pair} ({exit_reason} profit={profit} minutes={minutes} percent={last_candle['hapercent']}) but condition blocked") return (allow_to_sell) | (exit_reason == 'force_exit') | (exit_reason == 'stop_loss') def custom_stake_amount(self, pair: str, current_time: datetime, current_rate: float, proposed_stake: float, min_stake: float, max_stake: float, **kwargs) -> float: dataframe, _ = self.dp.get_analyzed_dataframe(pair=pair, timeframe=self.timeframe) current_candle = dataframe.iloc[-1].squeeze() adjusted_stake_amount = self.adjust_stake_amount(pair, current_candle) # print(f"{pair} adjusted_stake_amount{adjusted_stake_amount}") # Use default stake amount. return adjusted_stake_amount def custom_exit(self, pair: str, trade: Trade, current_time, current_rate, current_profit, **kwargs): dataframe, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe) last_candle = dataframe.iloc[-1].squeeze() last_candle_1h = dataframe.iloc[-13].squeeze() before_last_candle = dataframe.iloc[-2].squeeze() before_last_candle_2 = dataframe.iloc[-3].squeeze() before_last_candle_12 = dataframe.iloc[-13].squeeze() expected_profit = self.expectedProfit(pair, last_candle) # print(f"current_time={current_time} current_profit={current_profit} expected_profit={expected_profit}") max_touch_before = self.pairs[pair]['max_touch'] self.pairs[pair]['last_max'] = max(last_candle['close'], self.pairs[pair]['last_max']) self.pairs[pair]['last_min'] = min(last_candle['close'], self.pairs[pair]['last_min']) self.pairs[pair]['current_trade'] = trade count_of_buys = trade.nr_of_successful_entries profit = trade.calc_profit(current_rate) #round(current_profit * trade.stake_amount, 1) self.pairs[pair]['max_profit'] = max(self.pairs[pair]['max_profit'], profit) max_profit = self.pairs[pair]['max_profit'] baisse = 0 if profit > 0: baisse = 1 - (profit / max_profit) mx = max_profit / 5 self.pairs[pair]['count_of_buys'] = count_of_buys self.pairs[pair]['current_profit'] = profit dispo = round(self.wallets.get_available_stake_amount()) hours_since_first_buy = (current_time - trade.open_date_utc).seconds / 3600.0 days_since_first_buy = (current_time - trade.open_date_utc).days hours = (current_time - trade.date_last_filled_utc).total_seconds() / 3600.0 if hours % 4 == 0: self.log_trade( last_candle=last_candle, date=current_time, action="🔴 CURRENT" if self.pairs[pair]['stop'] or last_candle['stop_buying_1h'] else "🟢 CURRENT", dispo=dispo, pair=pair, rate=last_candle['close'], trade_type='', profit=round(profit, 2), buys=count_of_buys, stake=0 ) pair_name = self.getShortName(pair) if last_candle['max_rsi_24'] > 85 and profit > max(5, expected_profit) and (last_candle['hapercent'] < 0) and last_candle['sma60_deriv1'] < 0.05: self.pairs[pair]['force_sell'] = False self.pairs[pair]['force_buy'] = False #(self.pairs[pair]['count_of_buys'] - self.pairs[pair]['has_gain'] > 3) return str(count_of_buys) + '_' + 'Rsi85_' + pair_name + '_' + str(self.pairs[pair]['has_gain']) if self.pairs[pair]['force_sell']: self.pairs[pair]['force_sell'] = False self.pairs[pair]['force_buy'] = (self.pairs[pair]['count_of_buys'] - self.pairs[pair]['has_gain'] > 3) return str(count_of_buys) + '_' + 'Frc_' + pair_name + '_' + str(self.pairs[pair]['has_gain']) if profit > 0 and baisse > 0.30: self.pairs[pair]['force_sell'] = False self.pairs[pair]['force_buy'] = (self.pairs[pair]['count_of_buys'] - self.pairs[pair]['has_gain'] > 3) return str(count_of_buys) + '_' + 'B30_' + pair_name + '_' + str(self.pairs[pair]['has_gain']) # if max_profit > 0.5 * count_of_buys and baisse > 0.15: # self.pairs[pair]['force_sell'] = False # self.pairs[pair]['force_buy'] = (self.pairs[pair]['count_of_buys'] - self.pairs[pair]['has_gain'] > 3) # return str(count_of_buys) + '_' + 'B15_' + pair_name + '_' + str(self.pairs[pair]['has_gain']) if (last_candle['sma5_1h'] - before_last_candle_12['sma5_1h']) / last_candle['sma5_1h'] > 0.0002: return None factor = 1 if (self.getShortName(pair) == 'BTC'): factor = 0.5 # if baisse > 2 and baisse > factor * self.pairs[pair]['total_amount'] / 100: # self.pairs[pair]['force_sell'] = False # self.pairs[pair]['force_buy'] = (self.pairs[pair]['count_of_buys'] - self.pairs[pair]['has_gain'] > 3) # return 'Baisse_' + pair_name + '_' + str(count_of_buys) + '_' + str(self.pairs[pair]['has_gain']) # # if 1 <= count_of_buys <= 3: if last_candle['max_rsi_24'] > 75 and profit > expected_profit and (last_candle['hapercent'] < 0) and last_candle['sma60_deriv1'] < 0: self.pairs[pair]['force_sell'] = False return str(count_of_buys) + '_' + 'Rsi75_' + pair_name + '_' + str(self.pairs[pair]['has_gain']) self.pairs[pair]['max_touch'] = max(last_candle['close'], self.pairs[pair]['max_touch']) def getShortName(self, pair): return pair.replace("/USDT", '').replace("/USDC", '').replace("_USDC", '').replace("_USDT", '') def informative_pairs(self): # get access to all pairs available in whitelist. pairs = self.dp.current_whitelist() # informative_pairs = [(pair, '1d') for pair in pairs] informative_pairs += [(pair, '1h') for pair in pairs] return informative_pairs from typing import List def multi_step_interpolate(self, pct: float, thresholds: List[float], factors: List[float]) -> float: if pct <= thresholds[0]: return factors[0] if pct >= thresholds[-1]: return factors[-1] for i in range(1, len(thresholds)): if pct <= thresholds[i]: # interpolation linéaire entre thresholds[i-1] et thresholds[i] return factors[i - 1] + (pct - thresholds[i - 1]) * (factors[i] - factors[i - 1]) / ( thresholds[i] - thresholds[i - 1]) # Juste au cas où (devrait jamais arriver) return factors[-1] # def interpolate_factor(self, pct: float, start_pct: float = 5, end_pct: float = 30, # start_factor: float = 1.0, end_factor: float = 2.0) -> float: # if pct <= start_pct: # return start_factor # if pct >= end_pct: # return end_factor # # interpolation linéaire # return start_factor + (pct - start_pct) * (end_factor - start_factor) / (end_pct - start_pct) def log_trade(self, action, pair, date, trade_type=None, rate=None, dispo=None, profit=None, buys=None, stake=None, last_candle=None): # Afficher les colonnes une seule fois if self.config.get('runmode') == 'hyperopt' or self.dp.runmode.value in ('hyperopt'): return if self.columns_logged % 10 == 0: self.printLog( f"| {'Date':<16} | {'Action':<10} |{'Pair':<5}| {'Trade Type':<18} |{'Rate':>8} | {'Dispo':>6} | {'Profit':>8} " f"| {'Pct':>6} | {'max_touch':>11} | {'last_lost':>12} | {'last_max':>7}| {'last_max':>7}|{'Buys':>5}| {'Stake':>5} |" f"{'rsi':>6}|Distmax|s201d|s5_1d|s5_2d|s51h|s52h|smt1h|smt2h|tdc1d|tdc1h" ) self.printLineLog() df = pd.DataFrame.from_dict(self.pairs, orient='index') colonnes_a_exclure = ['last_candle', 'trade_info', 'last_date', 'last_count_of_buys', 'base_stake_amount', 'stop_buy'] df_filtered = df[df['count_of_buys'] > 0].drop(columns=colonnes_a_exclure) # df_filtered = df_filtered["first_buy", "last_max", "max_touch", "last_sell","last_buy", 'count_of_buys', 'current_profit'] print(df_filtered) self.columns_logged += 1 date = str(date)[:16] if date else "-" limit = None # if buys is not None: # limit = round(last_rate * (1 - self.fibo[buys] / 100), 4) rsi = '' rsi_pct = '' # if last_candle is not None: # if (not np.isnan(last_candle['rsi_1d'])) and (not np.isnan(last_candle['rsi_1h'])): # rsi = str(int(last_candle['rsi_1d'])) + " " + str(int(last_candle['rsi_1h'])) # if (not np.isnan(last_candle['rsi_pct_1d'])) and (not np.isnan(last_candle['rsi_pct_1h'])): # rsi_pct = str(int(10000 * last_candle['bb_mid_pct_1d'])) + " " + str( # int(last_candle['rsi_pct_1d'])) + " " + str(int(last_candle['rsi_pct_1h'])) # first_rate = self.percent_threshold.value # last_rate = self.threshold.value # action = self.color_line(action, action) sma5_1d = '' sma5_1h = '' sma5 = str(sma5_1d) + ' ' + str(sma5_1h) last_lost = self.getLastLost(last_candle, pair) if buys is None: buys = '' max_touch = '' pct_max = self.getPctFirstBuy(pair, last_candle) total_counts = str(buys) + '/' + str(sum(pair_data['count_of_buys'] for pair_data in self.pairs.values())) dist_max = '' color = GREEN if profit > 0 else RED color_sma24 = GREEN if last_candle['sma24_deriv1_1h'] > 0 else RED color_sma24_2 = GREEN if last_candle['sma24_deriv2_1h'] > 0 else RED color_sma5 = GREEN if last_candle['mid_smooth_5_deriv1_1h'] > 0 else RED color_sma5_2 = GREEN if last_candle['mid_smooth_5_deriv2_1h'] > 0 else RED color_sma5_1h = GREEN if last_candle['sma60_deriv1'] > 0 else RED color_sma5_2h = GREEN if last_candle['sma60_deriv2'] > 0 else RED color_smooth_1h = GREEN if last_candle['mid_smooth_1h_deriv1'] > 0 else RED color_smooth2_1h = GREEN if last_candle['mid_smooth_1h_deriv2'] > 0 else RED last_max = int(self.pairs[pair]['last_max']) if self.pairs[pair]['last_max'] > 1 else round( self.pairs[pair]['last_max'], 3) last_min = int(self.pairs[pair]['last_min']) if self.pairs[pair]['last_min'] > 1 else round( self.pairs[pair]['last_min'], 3) profit = str(profit) + '/' + str(round(self.pairs[pair]['max_profit'], 2)) # 🟢 Dérivée 1 > 0 et dérivée 2 > 0: tendance haussière qui s’accélère. # 🟡 Dérivée 1 > 0 et dérivée 2 < 0: tendance haussière qui ralentit → essoufflement potentiel. # 🔴 Dérivée 1 < 0 et dérivée 2 < 0: tendance baissière qui s’accélère. # 🟠 Dérivée 1 < 0 et dérivée 2 > 0: tendance baissière qui ralentit → possible bottom. self.printLog( f"| {date:<16} |{action:<10} | {pair[0:3]:<3} | {trade_type or '-':<18} |{rate or '-':>9}| {dispo or '-':>6} " f"|{color}{profit or '-':>10}{RESET}| {pct_max or '-':>6} | {round(self.pairs[pair]['max_touch'], 2) or '-':>11} | {last_lost or '-':>12} " f"| {last_max or '-':>7} | {last_min or '-':>7} |{total_counts or '-':>5}|{stake or '-':>7}" # f"|{round(last_candle['mid_smooth_24_deriv1'],3) or '-':>6}|{round(last_candle['mid_smooth_1h_deriv1'],3) or '-':>6}|{round(last_candle['mid_smooth_deriv1_1d'],3) or '-' :>6}|" # f"{round(last_candle['mid_smooth_24_deriv2'],3) or '-' :>6}|{round(last_candle['mid_smooth_1h_deriv2'],3) or '-':>6}|{round(last_candle['mid_smooth_deriv2_1d'],3) or '-':>6}|" f"{round(last_candle['max_rsi_24'], 1) or '-' :>6}|" f"{dist_max:>7}|{color_sma24}{round(last_candle['sma24_deriv1_1h'], 2):>5}{RESET}" f"|{color_sma5}{round(last_candle['mid_smooth_5_deriv1_1h'], 2):>5}{RESET}|{color_sma5_2}{round(last_candle['mid_smooth_5_deriv2_1h'], 2):>5}{RESET}" f"|{color_sma5_1h}{round(last_candle['sma60_deriv1'], 2):>5}{RESET}|{color_sma5_2h}{round(last_candle['sma60_deriv2'], 2):>5}{RESET}" f"|{color_smooth_1h}{round(last_candle['mid_smooth_1h_deriv1'], 2):>5}{RESET}|{color_smooth2_1h}{round(last_candle['mid_smooth_1h_deriv2'], 2):>5}{RESET}" ) def getLastLost(self, last_candle, pair): last_lost = round((last_candle['close'] - self.pairs[pair]['max_touch']) / self.pairs[pair]['max_touch'], 3) return last_lost def printLineLog(self): # f"sum1h|sum1d|Tdc|Tdh|Tdd| drv1 |drv_1h|drv_1d|" self.printLog( f"+{'-' * 18}+{'-' * 12}+{'-' * 5}+{'-' * 20}+{'-' * 9}+{'-' * 8}+{'-' * 12}+{'-' * 8}+{'-' * 13}+{'-' * 14}+{'-' * 9}{'-' * 9}+{'-' * 5}+{'-' * 7}+" f"+{'-' * 6}+{'-' * 7}+{'-' * 5}+{'-' * 5}+{'-' * 5}+{'-' * 5}+{'-' * 5}+{'-' * 5}+" ) def printLog(self, str): if self.config.get('runmode') == 'hyperopt' or self.dp.runmode.value in ('hyperopt'): return; if not self.dp.runmode.value in ('backtest', 'hyperopt', 'lookahead-analysis'): logger.info(str) else: if not self.dp.runmode.value in ('hyperopt'): print(str) def add_tendency_column(self, dataframe: pd.DataFrame, name: str, suffixe: str = '', eps: float = 1e-3, d1_lim_inf: float = -0.01, d1_lim_sup: float = 0.01) -> pd.DataFrame: """ Ajoute une colonne 'tendency' basée sur les dérivées 1 et 2 lissées et normalisées. eps permet de définir un seuil proche de zéro. suffixe permet de gérer plusieurs indicateurs. """ def tag_by_derivatives(row): d1 = row[f"{name}{suffixe}_deriv1"] d2 = row[f"{name}{suffixe}_deriv2"] # On considère les petites valeurs comme zéro if abs(d1) < eps: return 0 # Palier / neutre if d1 > d1_lim_sup: return 2 if d2 > eps else 1 # Acceleration Hausse / Ralentissement Hausse if d1 < d1_lim_inf: return -2 if d2 < -eps else -1 # Acceleration Baisse / Ralentissement Baisse if abs(d1) < eps: return 'DH' if d2 > eps else 'DB' # Depart Hausse / Depart Baisse return 'Mid' print(f"{name}_tdc{suffixe}") dataframe[f"{name}_tdc{suffixe}"] = dataframe.apply(tag_by_derivatives, axis=1) return dataframe # def add_tendency_column(self, dataframe: pd.DataFrame, name, suffixe='') -> pd.DataFrame: # def tag_by_derivatives(row): # d1 = row[f"{name}{suffixe}_deriv1"] # d2 = row[f"{name}{suffixe}_deriv2"] # d1_lim_inf = -0.01 # d1_lim_sup = 0.01 # if d1 >= d1_lim_inf and d1 <= d1_lim_sup: # and d2 >= d2_lim_inf and d2 <= d2_lim_sup: # return 0 # Palier # if d1 == 0.0: # return 'DH' if d2 > 0 else 'DB' # Depart Hausse / Départ Baisse # if d1 > d1_lim_sup: # return 2 if d2 > 0 else 1 # Acceleration Hausse / Ralentissement Hausse # if d1 < d1_lim_inf: # return -2 if d2 < 0 else -1 # Accéleration Baisse / Ralentissement Baisse # return 'Mid' # # dataframe[f"tendency{suffixe}"] = dataframe.apply(tag_by_derivatives, axis=1) # return dataframe def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame: # Add all ta features pair = metadata['pair'] short_pair = self.getShortName(pair) self.path = f"user_data/plots/{short_pair}/" dataframe = self.populateDataframe(dataframe, timeframe='5m') ################### INFORMATIVE 1h informative = self.dp.get_pair_dataframe(pair=metadata['pair'], timeframe="1h") informative = self.populateDataframe(informative, timeframe='1h') informative = self.calculateRegression(informative, 'mid', lookback=5) # # TENSOR FLOW # self.model_indicators = self.listUsableColumns(informative) # if self.dp.runmode.value in ('backtest'): # self.trainTensorFlow(informative, future_steps = self.future_steps) # # self.predictTensorFlow(informative) # # if self.dp.runmode.value in ('backtest'): # self.kerasGenerateGraphs(informative) informative['stop_buying_deb'] = ((informative['max_rsi_24'] > self.rsi_deb_protect.value) & (informative['sma24_deriv1'] < self.sma24_deriv1_deb_protect.value) ) informative['stop_buying_end'] = ((informative['max_rsi_24'] < self.rsi_end_protect.value) & (informative['sma24_deriv1'] > self.sma24_deriv1_end_protect.value) ) latched = np.zeros(len(informative), dtype=bool) for i in range(1, len(informative)): if informative['stop_buying_deb'].iloc[i]: latched[i] = True elif informative['stop_buying_end'].iloc[i]: latched[i] = False else: latched[i] = latched[i - 1] informative['stop_buying'] = latched dataframe = merge_informative_pair(dataframe, informative, self.timeframe, "1h", ffill=True) # ################### INFORMATIVE 1d # informative = self.dp.get_pair_dataframe(pair=metadata['pair'], timeframe="1d") # informative = self.populateDataframe(informative, timeframe='1d') # # informative = self.calculateRegression(informative, 'mid', lookback=15) # dataframe = merge_informative_pair(dataframe, informative, self.timeframe, "1d", ffill=True) dataframe['last_price'] = dataframe['close'] dataframe['first_price'] = dataframe['close'] if self.dp: if self.dp.runmode.value in ('live', 'dry_run'): self.getOpenTrades() for trade in self.trades: if trade.pair != pair: continue filled_buys = trade.select_filled_orders('buy') count = 0 amount = 0 for buy in filled_buys: if count == 0: dataframe['first_price'] = buy.price self.pairs[pair]['first_buy'] = buy.price self.pairs[pair]['first_amount'] = buy.price * buy.filled # dataframe['close01'] = buy.price * 1.01 # Order(id=2396, trade=1019, order_id=29870026652, side=buy, filled=0.00078, price=63921.01, # status=closed, date=2024-08-26 02:20:11) dataframe['last_price'] = buy.price self.pairs[pair]['last_buy'] = buy.price count = count + 1 amount += buy.price * buy.filled # dataframe['mid_price'] = (dataframe['last_price'] + dataframe['first_price']) / 2 count_buys = count # dataframe['limit'] = dataframe['last_price'] * (1 - self.baisse[count] / 100) self.pairs[pair]['total_amount'] = amount # dataframe['mid_smooth_tag'] = qtpylib.crossed_below(dataframe['mid_smooth_24_deriv1'], dataframe['mid_smooth_deriv2_24']) # =============================== # lissage des valeurs horaires dataframe['mid_smooth_1h'] = dataframe['mid'].rolling(window=6).mean() dataframe["mid_smooth_1h_deriv1"] = 100 * dataframe["mid_smooth_1h"].diff().rolling(window=6).mean() / \ dataframe['mid_smooth_1h'] dataframe["mid_smooth_1h_deriv2"] = 100 * dataframe["mid_smooth_1h_deriv1"].diff().rolling(window=6).mean() dataframe['mid_smooth_5h'] = talib.EMA(dataframe, timeperiod=60) # dataframe['mid'].rolling(window=60).mean() dataframe["mid_smooth_5h_deriv1"] = 100 * dataframe["mid_smooth_5h"].diff().rolling(window=60).mean() / \ dataframe['mid_smooth_5h'] dataframe["mid_smooth_5h_deriv2"] = 100 * dataframe["mid_smooth_5h_deriv1"].diff().rolling(window=60).mean() dataframe = self.calculateRegression(dataframe, 'mid', lookback=10, future_steps=10, model_type="poly") dataframe = self.calculateRegression(dataframe, 'sma24', lookback=12, future_steps=12) # dataframe["ms-10"] = dataframe[self.indicator_target].shift(10) # dataframe["ms-5"] = dataframe[self.indicator_target].shift(5) # dataframe["ms-4"] = dataframe[self.indicator_target].shift(4) # dataframe["ms-3"] = dataframe[self.indicator_target].shift(3) # dataframe["ms-2"] = dataframe[self.indicator_target].shift(2) # dataframe["ms-1"] = dataframe[self.indicator_target].shift(1) # dataframe["ms-0"] = dataframe[self.indicator_target] # dataframe["ms+10"] = dataframe["mid_smooth_24"].shift(-11) self.model_indicators = self.listUsableColumns(dataframe) # # Quantile # self.add_future_quantiles( # dataframe, # indic="mid", # lookback=40, # future_steps=5 # ) # TENSOR FLOW if self.dp.runmode.value in ('backtest'): self.trainTensorFlow(dataframe, future_steps = self.future_steps) self.predictTensorFlow(dataframe) if self.dp.runmode.value in ('backtest'): self.kerasGenerateGraphs(dataframe) # SKLEARN # if self.dp.runmode.value in ('backtest'): # self.trainModel(dataframe, metadata) # short_pair = self.getShortName(pair) # self.model = joblib.load(f"{short_pair}_rf_model.pkl") # # # Préparer les features pour la prédiction # features = dataframe[self.model_indicators].fillna(0) # # # Prédiction : probabilité que le prix monte # # probs = self.model.predict_proba(features)[:, 1] # probs = self.model.predict(features) # # # Sauvegarder la probabilité pour l’analyse # dataframe['ml_prob'] = probs # # self.inspect_model(self.model) return dataframe def trainModel(self, dataframe: DataFrame, metadata: dict): pair = self.getShortName(metadata['pair']) pd.set_option('display.max_rows', None) pd.set_option('display.max_columns', None) pd.set_option("display.width", 200) os.makedirs(self.path, exist_ok=True) df = dataframe[self.model_indicators].copy() # Corrélations des colonnes corr = df.corr(numeric_only=True) print("Corrélation des colonnes") print(corr) # 3️⃣ Créer la cible : 1 si le prix monte dans les prochaines bougies # df['target'] = (df['sma24'].shift(-24) > df['sma24']).astype(int) df['target'] = dataframe[self.indicator_target].shift(-24) # > df['sma24'] * 1.003).astype(int) df['target'] = df['target'].fillna(0) #.astype(int) # Corrélations triées par importance avec une colonne cible target_corr = df.corr(numeric_only=True)["target"].sort_values(ascending=False) print("Corrélations triées par importance avec une colonne cible") print(target_corr) # Corrélations triées par importance avec une colonne cible corr = df.corr(numeric_only=True) corr_unstacked = ( corr.unstack() .reset_index() .rename(columns={"level_0": "col1", "level_1": "col2", 0: "corr"}) ) # Supprimer les doublons col1/col2 inversés et soi-même corr_unstacked = corr_unstacked[corr_unstacked["col1"] < corr_unstacked["col2"]] # Trier par valeur absolue de corrélation corr_sorted = corr_unstacked.reindex(corr_unstacked["corr"].abs().sort_values(ascending=False).index) print("Trier par valeur absolue de corrélation") print(corr_sorted.head(20)) # --- Calcul de la corrélation --- corr = df.corr(numeric_only=True) # évite les colonnes non numériques corr = corr * 100 # passage en pourcentage # --- Masque pour n’afficher que le triangle supérieur (optionnel) --- mask = np.triu(np.ones_like(corr, dtype=bool)) # --- Création de la figure --- fig, ax = plt.subplots(figsize=(96, 36)) # --- Heatmap avec un effet “température” --- sns.heatmap( corr, mask=mask, cmap="coolwarm", # palette bleu → rouge center=0, # 0 au centre annot=True, # affiche les valeurs dans chaque case fmt=".0f", # format entier (pas de décimale) cbar_kws={"label": "Corrélation (%)"}, # légende à droite linewidths=0.5, # petites lignes entre les cases ax=ax ) # --- Personnalisation --- ax.set_title("Matrice de corrélation (en %)", fontsize=20, pad=20) plt.xticks(rotation=45, ha="right") plt.yticks(rotation=0) # --- Sauvegarde --- output_path = f"{self.path}/Matrice_de_correlation_temperature.png" plt.savefig(output_path, bbox_inches="tight", dpi=150) plt.close(fig) print(f"✅ Matrice enregistrée : {output_path}") # Nettoyage df = df.dropna() X = df[self.model_indicators] y = df['target'] # ta colonne cible binaire ou numérique print(self.feature_auc_scores(X, y)) # 4️⃣ Split train/test X = df[self.model_indicators] y = df['target'] # Séparation temporelle (train = 80 %, valid = 20 %) X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, shuffle=False) print("NaN per column:") print(X_train.isna().sum().sort_values(ascending=False).head(20)) # Nettoyage des valeurs invalides selector = VarianceThreshold(threshold=0.0001) selector.fit(X_train) selected = X_train.columns[selector.get_support()] print("Colonnes conservées :", list(selected)) # 1️⃣ Entraîne ton modèle LGBM normal # train_model = LGBMRegressor( # objective='regression', # metric='rmse', # tu peux aussi tester 'mae' # n_estimators=300, # learning_rate=0.05, # max_depth=7, # subsample=0.8, # colsample_bytree=0.8, # random_state=42 # ) # train_model.fit(X_train, y_train) train_model, selected_features = self.optuna(self.path, X_train, X_test, y_train, y_test) print("Features retenues :", list(selected_features)) # # 2️⃣ Sélection des features AVANT calibration # sfm = SelectFromModel(train_model, threshold="median", prefit=True) # selected_features = X_train.columns[sfm.get_support()] # print(selected_features) train_model.fit(X_train, y_train) # Importances importances = pd.DataFrame({ "feature": train_model.feature_name_, "importance": train_model.feature_importances_ }).sort_values("importance", ascending=False) print("\n===== 🔍 IMPORTANCE DES FEATURES =====") print(importances) # 6️⃣ Évaluer la précision (facultatif) preds = train_model.predict(X_test) mse = mean_squared_error(y_test, preds) rmse = np.sqrt(mse) r2 = r2_score(y_test, preds) print(f"RMSE: {rmse:.5f} | R²: {r2:.3f}") # acc = accuracy_score(y_test, preds) # print(f"Accuracy: {acc:.3f}") # 7️⃣ Sauvegarde du modèle joblib.dump(train_model, f"{pair}_rf_model.pkl") print(f"✅ Modèle sauvegardé sous {pair}_rf_model.pkl") # # Quantile # dataframe = self.add_future_quantiles( # df, # indic="mid", # lookback=40, # future_steps=5 # ) self.analyze_model(pair, train_model, X_train, X_test, y_train, y_test) def listUsableColumns(self, dataframe): # Étape 1 : sélectionner numériques numeric_cols = dataframe.select_dtypes(include=['int64', 'float64']).columns # Étape 2 : enlever constantes usable_cols = [c for c in numeric_cols if dataframe[c].nunique() > 1 and not c.endswith("_state") and not c.endswith("_1d") # and not c.endswith("_1h") and not c.endswith("_count") # and not c.startswith("open") and not c.startswith("close") # and not c.startswith("low") and not c.startswith("high") # and not c.startswith("haopen") and not c.startswith("haclose") # and not c.startswith("bb_lower") and not c.startswith("bb_upper") # and not c.startswith("bb_middle") and not c.endswith("_class") and not c.endswith("_price") and not c.startswith('stop_buying')] # Étape 3 : remplacer inf et NaN par 0 dataframe[usable_cols] = dataframe[usable_cols].replace([np.inf, -np.inf], 0).fillna(0) print("Colonnes utilisables pour le modèle :") print(usable_cols) self.model_indicators = usable_cols # self.model_indicators = [ # 'volume', 'hapercent', 'mid', 'percent', 'percent3', 'percent12', # 'percent24', # 'sma5', 'sma5_dist', 'sma5_deriv1', 'sma5_deriv2', 'sma12', 'sma12_dist', # 'sma12_deriv1', 'sma12_deriv2', 'sma24', 'sma24_dist', 'sma24_deriv1', 'sma24_deriv2', # # 'sma48', 'sma48_dist', 'sma48_deriv1', 'sma48_deriv2', 'sma60', 'sma60_dist', # # 'sma60_deriv1', 'sma60_deriv2', 'mid_smooth_3', 'mid_smooth_3_dist', # # 'mid_smooth_3_deriv1', 'mid_smooth_3_deriv2', 'mid_smooth_5', 'mid_smooth_5_dist', # # 'mid_smooth_5_deriv1', 'mid_smooth_5_deriv2', 'mid_smooth_12', 'mid_smooth_12_dist', # # 'mid_smooth_12_deriv1', 'mid_smooth_12_deriv2', 'mid_smooth_24', 'mid_smooth_24_dist', # # 'mid_smooth_24_deriv1', 'mid_smooth_24_deriv2', 'rsi', 'max_rsi_12', 'max_rsi_24', # 'rsi_dist', 'rsi_deriv1', 'rsi_deriv2', 'max12', 'min12', 'max60', 'min60', # 'min_max_60', 'bb_percent', 'bb_width', 'macd', 'macdsignal', 'macdhist', 'slope', # 'slope_smooth', 'atr', 'atr_norm', 'adx', 'obv', 'vol_24', # # 'down_count', 'up_count', # # 'down_pct', 'up_pct', 'rsi_slope', 'adx_change', 'volatility_ratio', 'rsi_diff', # # 'slope_ratio', 'volume_sma_deriv', 'volume_dist', 'volume_deriv1', 'volume_deriv2', # # 'slope_norm', 'mid_smooth_1h_deriv1', 'mid_smooth_1h_deriv2', 'mid_smooth_5h', # # 'mid_smooth_5h_deriv1', 'mid_smooth_5h_deriv2', 'mid_future_pred_cons', # # 'sma24_future_pred_cons' # ] return self.model_indicators def inspect_model(self, model): """ Affiche les informations d'un modèle ML déjà entraîné. Compatible avec scikit-learn, xgboost, lightgbm, catboost... """ print("===== 🔍 INFORMATIONS DU MODÈLE =====") # Type de modèle print(f"Type : {type(model).__name__}") print(f"Module : {model.__class__.__module__}") # Hyperparamètres if hasattr(model, "get_params"): params = model.get_params() print(f"\n===== ⚙️ HYPERPARAMÈTRES ({len(params)}) =====") for k, v in params.items(): print(f"{k}: {v}") # Nombre d’estimateurs if hasattr(model, "n_estimators"): print(f"\nNombre d’estimateurs : {model.n_estimators}") # Importance des features if hasattr(model, "feature_importances_"): print("\n===== 📊 IMPORTANCE DES FEATURES =====") # Correction ici : feature_names = getattr(model, "feature_names_in_", None) if isinstance(feature_names, np.ndarray): feature_names = feature_names.tolist() elif feature_names is None: feature_names = [f"feature_{i}" for i in range(len(model.feature_importances_))] fi = pd.DataFrame({ "feature": feature_names, "importance": model.feature_importances_ }).sort_values(by="importance", ascending=False) print(fi) # Coefficients (modèles linéaires) if hasattr(model, "coef_"): print("\n===== ➗ COEFFICIENTS =====") coef = np.array(model.coef_) if coef.ndim == 1: for i, c in enumerate(coef): print(f"Feature {i}: {c:.6f}") else: print(coef) # Intercept if hasattr(model, "intercept_"): print("\nIntercept :", model.intercept_) # Classes connues if hasattr(model, "classes_"): print("\n===== 🎯 CLASSES =====") print(model.classes_) # Scores internes for attr in ["best_score_", "best_iteration_", "best_ntree_limit", "score_"]: if hasattr(model, attr): print(f"\n{attr} = {getattr(model, attr)}") # Méthodes disponibles print("\n===== 🧩 MÉTHODES DISPONIBLES =====") methods = [m for m, _ in inspect.getmembers(model, predicate=inspect.ismethod)] print(", ".join(methods[:15]) + ("..." if len(methods) > 15 else "")) print("\n===== ✅ FIN DE L’INSPECTION =====") def analyze_model(self, pair, model, X_train, X_test, y_train, y_test): """ Analyse complète d'un modèle ML supervisé (classification binaire). Affiche performances, importance des features, matrices, seuils, etc. """ output_dir = f"user_data/plots/{pair}/" os.makedirs(output_dir, exist_ok=True) # ---- Importance des features ---- if hasattr(model, "feature_importances_"): print("\n===== 🔍 IMPORTANCE DES FEATURES =====") importance = pd.DataFrame({ "feature": X_train.columns, "importance": model.feature_importances_ }).sort_values(by="importance", ascending=False) print(importance) top_n = 20 importance = importance.head(top_n) # Crée une figure plus grande fig, ax = plt.subplots(figsize=(24, 8)) # largeur=24 pouces, hauteur=8 pouces # Trace le bar plot sur cet axe importance.plot.bar(x="feature", y="importance", legend=False, ax=ax) # Tourner les labels pour plus de lisibilité ax.set_xticklabels(ax.get_xticklabels(), rotation=45, ha='right') plt.title("Importance des features") # plt.show() plt.savefig(os.path.join(output_dir, "Importance des features.png"), bbox_inches="tight") plt.close() # ---- Arbre de décision (extrait) ---- if hasattr(model, "estimators_"): print("\n===== 🌳 EXTRAIT D’UN ARBRE =====") print(export_text(model.estimators_[0], feature_names=list(X_train.columns))[:800]) # --- Après l'entraînement du modèle --- preds = model.predict(X_test) # --- Évaluation --- mse = mean_squared_error(y_test, preds) rmse = np.sqrt(mse) r2 = r2_score(y_test, preds) print(f"RMSE: {rmse:.5f} | R²: {r2:.3f}") # --- Création du dossier de sortie --- os.makedirs(output_dir, exist_ok=True) # --- Graphique prédiction vs réel --- plt.figure(figsize=(8, 8)) plt.scatter(y_test, preds, alpha=0.4, s=15) plt.xlabel("Valeurs réelles", fontsize=12) plt.ylabel("Valeurs prédites", fontsize=12) plt.title(f"LightGBM Régression — Prédiction vs Réel\nRMSE={rmse:.5f} | R²={r2:.3f}", fontsize=14) plt.plot( [y_test.min(), y_test.max()], [y_test.min(), y_test.max()], 'r--', linewidth=1, label="Ligne idéale" ) plt.legend() # --- Sauvegarde --- plot_path = os.path.join(output_dir, "LightGBM_regression_pred_vs_real.png") plt.savefig(plot_path, bbox_inches="tight", dpi=200) plt.close() self.plot_pred_vs_real_filtered(model, X_test, y_test, preds, output_dir) print(f"✅ Graphique sauvegardé : {plot_path}") # ax = lgb.plot_tree(model, tree_index=0, figsize=(30, 20), show_info=["split_gain", "internal_value", "internal_count"]) # plt.title("Arbre de décision n°0") # plt.savefig(os.path.join(output_dir, "lgbm_tree_0.png"), bbox_inches="tight") # plt.close() for i in range(5): ax = lgb.plot_tree(model, tree_index=i, figsize=(20, 12)) plt.title(f"Arbre {i}") plt.savefig(os.path.join(output_dir, f"lgbm_tree_{i}.png"), bbox_inches="tight") plt.close() ax = lgb.plot_tree(model, figsize=(40, 20)) plt.title("Vue globale du modèle LGBM") plt.savefig(os.path.join(output_dir, "lgbm_all_trees.png"), bbox_inches="tight") plt.close() # X_test = np.linspace(0, 10, 1000).reshape(-1, 1) y_pred = model.predict(X_test) self.graphFonctionApprise(output_dir, X_test, y_test, y_pred) self.graphFonctionAppriseFeature(output_dir, X_test, y_test, y_pred) # ============================================================================== ax = lgb.plot_importance(model, max_num_features=30, figsize=(12, 6)) plt.title("Importance des features - LGBM") plt.savefig(os.path.join(output_dir, "lgbm_feature_importance.png"), bbox_inches="tight") plt.close() corr = X_train.corr() * 100 # en pourcentage plt.figure(figsize=(20, 16)) sns.heatmap(corr, cmap="coolwarm", center=0, annot=False, fmt=".1f", cbar_kws={'label': 'Corrélation (%)'}) plt.title("Matrice de corrélation (%)") plt.savefig(os.path.join(output_dir, "correlation_matrix.png"), bbox_inches="tight") plt.close() plt.figure(figsize=(10, 6)) plt.scatter(y_test, model.predict(X_test), alpha=0.5) plt.xlabel("Valeurs réelles") plt.ylabel("Prédictions du modèle") plt.title("Comparaison y_test vs y_pred") plt.savefig(os.path.join(output_dir, "ytest_vs_ypred.png"), bbox_inches="tight") plt.close() print("\n===== ✅ FIN DE L’ANALYSE =====") def plot_pred_vs_real_filtered(self, model, X_test, y_test, preds, output_dir, top_n=5): """ Affiche le graphique prédiction vs réel pour les N features les plus importantes. """ # --- 1️⃣ Extraire les features les plus importantes --- importance_df = pd.DataFrame({ "feature": X_test.columns, "importance": model.feature_importances_ }).sort_values(by="importance", ascending=False) top_features = importance_df.head(top_n)["feature"].tolist() print(f"Top {top_n} features: {top_features}") # --- 2️⃣ Créer un masque pour ne garder que les lignes où au moins une des top features varie fortement --- X_top = X_test[top_features] # Optionnel : filtrer les points atypiques pour lisser le nuage mask = np.all(np.abs((X_top - X_top.mean()) / X_top.std()) < 3, axis=1) X_filtered = X_top[mask] y_filtered = y_test[mask] preds_filtered = preds[mask] # --- 3️⃣ Tracer --- plt.figure(figsize=(8, 8)) plt.scatter(y_filtered, preds_filtered, alpha=0.4, s=15, c='blue', label=f"Top {top_n} features") plt.xlabel("Valeurs réelles", fontsize=12) plt.ylabel("Valeurs prédites", fontsize=12) plt.title(f"LightGBM Régression — Prédiction vs Réel (filtré sur top {top_n} features)", fontsize=14) plt.plot( [y_filtered.min(), y_filtered.max()], [y_filtered.min(), y_filtered.max()], 'r--', linewidth=1, label="Ligne idéale" ) plt.legend() plt.grid(True) out_path = f"{output_dir}/lgbm_pred_vs_real_top{top_n}.png" plt.savefig(out_path, bbox_inches="tight") plt.close() def plot_threshold_analysis(self, y_true, y_proba, step=0.05, save_path=None): """ Affiche la précision, le rappel et le F1-score selon le seuil de décision. y_true : labels réels (0 ou 1) y_proba : probabilités prédites (P(hausse)) step : pas entre les seuils testés save_path : si renseigné, enregistre l'image au lieu d'afficher """ # Le graphique généré affichera trois courbes : # # 🔵 Precision — la fiabilité de tes signaux haussiers. # 🟢 Recall — la proportion de hausses que ton modèle détecte. # 🟣 F1-score — le compromis optimal entre les deux. thresholds = np.arange(0, 1.01, step) precisions, recalls, f1s = [], [], [] for thr in thresholds: preds = (y_proba >= thr).astype(int) precisions.append(precision_score(y_true, preds)) recalls.append(recall_score(y_true, preds)) f1s.append(f1_score(y_true, preds)) plt.figure(figsize=(10, 6)) plt.plot(thresholds, precisions, label="Precision", linewidth=2) plt.plot(thresholds, recalls, label="Recall", linewidth=2) plt.plot(thresholds, f1s, label="F1-score", linewidth=2, linestyle="--") plt.axvline(0.5, color='gray', linestyle=':', label="Seuil 0.5") plt.title("📊 Performance selon le seuil de probabilité", fontsize=14) plt.xlabel("Seuil de décision (threshold)") plt.ylabel("Score") plt.legend() plt.grid(True, alpha=0.3) if save_path: plt.savefig(save_path, bbox_inches='tight') print(f"✅ Graphique enregistré : {save_path}") else: plt.show() # # ============================= # # Exemple d’utilisation : # # ============================= # if __name__ == "__main__": # # Exemple : chargement d’un modèle et test # import joblib # # model = joblib.load("/media/Home/home/souti/freqtrade/user_data/strategies/tools/sklearn/model.pkl") # data = np.load("/media/Home/home/souti/freqtrade/user_data/strategies/tools/sklearn/test_data.npz") # X_test, y_test = data["X"], data["y"] # # y_proba = model.predict_proba(X_test)[:, 1] # # # Trace ou enregistre le graphique # plot_threshold_analysis(y_test, y_proba, step=0.05, # save_path="/media/Home/home/souti/freqtrade/user_data/strategies/tools/sklearn/threshold_analysis.png") def populateDataframe(self, dataframe, timeframe='5m'): dataframe = dataframe.copy() heikinashi = qtpylib.heikinashi(dataframe) dataframe['haopen'] = heikinashi['open'] dataframe['haclose'] = heikinashi['close'] dataframe['hapercent'] = (dataframe['haclose'] - dataframe['haopen']) / dataframe['haclose'] dataframe['mid'] = dataframe['haopen'] + (dataframe['haclose'] - dataframe['haopen']) / 2 dataframe["percent"] = dataframe['close'].pct_change() dataframe["percent3"] = dataframe['close'].pct_change(3).rolling(3).mean() dataframe["percent12"] = dataframe['close'].pct_change(12).rolling(12).mean() dataframe["percent24"] = dataframe['close'].pct_change(24).rolling(24).mean() # if self.dp.runmode.value in ('backtest'): # dataframe['futur_percent'] = 100 * (dataframe['close'].shift(-1) - dataframe['close']) / dataframe['close'] dataframe['sma5'] = dataframe['mid'].ewm(span=5, adjust=False).mean() #dataframe["mid"].rolling(window=5).mean() self.calculeDerivees(dataframe, 'sma5', timeframe=timeframe, ema_period=5) dataframe['sma12'] = dataframe['mid'].ewm(span=12, adjust=False).mean() #dataframe["mid"].rolling(window=12).mean() self.calculeDerivees(dataframe, 'sma12', timeframe=timeframe, ema_period=12) dataframe['sma24'] = dataframe['mid'].ewm(span=24, adjust=False).mean() #dataframe["mid"].rolling(window=24).mean() self.calculeDerivees(dataframe, 'sma24', timeframe=timeframe, ema_period=24) dataframe['sma48'] = dataframe['mid'].ewm(span=48, adjust=False).mean() #dataframe["mid"].rolling(window=48).mean() self.calculeDerivees(dataframe, 'sma48', timeframe=timeframe, ema_period=48) dataframe['sma60'] = dataframe['mid'].ewm(span=60, adjust=False).mean() #dataframe["mid"].rolling(window=60).mean() self.calculeDerivees(dataframe, 'sma60', timeframe=timeframe, ema_period=60) dataframe = self.calculateDerivation(dataframe, window=3, suffixe="_3",timeframe=timeframe) dataframe = self.calculateDerivation(dataframe, window=5, suffixe="_5",timeframe=timeframe) dataframe = self.calculateDerivation(dataframe, window=12, suffixe="_12",timeframe=timeframe) dataframe = self.calculateDerivation(dataframe, window=24, suffixe="_24", timeframe=timeframe) # print(metadata['pair']) dataframe['rsi'] = talib.RSI(dataframe['close'], timeperiod=14) dataframe['max_rsi_12'] = talib.MAX(dataframe['rsi'], timeperiod=12) dataframe['max_rsi_24'] = talib.MAX(dataframe['rsi'], timeperiod=24) self.calculeDerivees(dataframe, 'rsi', timeframe=timeframe, ema_period=12) dataframe['max12'] = talib.MAX(dataframe['close'], timeperiod=12) dataframe['min12'] = talib.MIN(dataframe['close'], timeperiod=12) dataframe['max60'] = talib.MAX(dataframe['close'], timeperiod=60) dataframe['min60'] = talib.MIN(dataframe['close'], timeperiod=60) dataframe['min_max_60'] = ((dataframe['max60'] - dataframe['close']) / dataframe['min60']) # dataframe['min36'] = talib.MIN(dataframe['close'], timeperiod=36) # dataframe['max36'] = talib.MAX(dataframe['close'], timeperiod=36) # dataframe['pct36'] = 100 * (dataframe['max36'] - dataframe['min36']) / dataframe['min36'] # dataframe['maxpct36'] = talib.MAX(dataframe['pct36'], timeperiod=36) # Bollinger Bands bollinger = qtpylib.bollinger_bands(qtpylib.typical_price(dataframe), window=20, stds=2) dataframe['bb_lowerband'] = bollinger['lower'] dataframe['bb_middleband'] = bollinger['mid'] dataframe['bb_upperband'] = bollinger['upper'] dataframe["bb_percent"] = ( (dataframe["close"] - dataframe["bb_lowerband"]) / (dataframe["bb_upperband"] - dataframe["bb_lowerband"]) ) dataframe["bb_width"] = (dataframe["bb_upperband"] - dataframe["bb_lowerband"]) / dataframe["sma5"] # dataframe["bb_width"] = ( # (dataframe["bb_upperband"] - dataframe["bb_lowerband"]) / dataframe["bb_middleband"] # ) # Calcul MACD macd, macdsignal, macdhist = talib.MACD( dataframe['close'], fastperiod=12, slowperiod=26, signalperiod=9 ) # | Nom | Formule / définition | Signification | # | ---------------------------- | ------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | # | **MACD** (`macd`) | `EMA_fast - EMA_slow` (ex : 12-26 périodes) | Montre l’écart entre la moyenne courte et la moyenne longue.
- Positive → tendance haussière
- Négative → tendance baissière | # | **Signal** (`macdsignal`) | `EMA_9(MACD)` | Sert de ligne de **signal de déclenchement**.
- Croisement du MACD au-dessus → signal d’achat
- Croisement du MACD en dessous → signal de vente | # | **Histogramme** (`macdhist`) | `MACD - Signal` | Montre la **force et l’accélération** de la tendance.
- Positif et croissant → tendance haussière qui s’accélère
- Positif mais décroissant → ralentissement de la hausse
- Négatif et décroissant → baisse qui s’accélère
- Négatif mais croissant → ralentissement de la baisse | # Ajouter dans le dataframe dataframe['macd'] = macd dataframe['macdsignal'] = macdsignal dataframe['macdhist'] = macdhist # Regarde dans le futur # # --- Rendre relatif sur chaque série (-1 → 1) --- # for col in ['macd', 'macdsignal', 'macdhist']: # series = dataframe[col] # valid = series[~np.isnan(series)] # ignorer NaN # min_val = valid.min() # max_val = valid.max() # span = max_val - min_val if max_val != min_val else 1 # dataframe[f'{col}_rel'] = 2 * ((series - min_val) / span) - 1 # # dataframe['tdc_macd'] = self.macd_tendance_int( # dataframe, # macd_col='macd_rel', # signal_col='macdsignal_rel', # hist_col='macdhist_rel' # ) # --- pente brute --- dataframe['slope'] = dataframe['sma24'].diff() # --- lissage EMA --- dataframe['slope_smooth'] = dataframe['slope'].ewm(span=10, adjust=False).mean() # --- Volatilité normalisée --- dataframe['atr'] = ta.volatility.AverageTrueRange( high=dataframe['high'], low=dataframe['low'], close=dataframe['close'], window=14 ).average_true_range() dataframe['atr_norm'] = dataframe['atr'] / dataframe['close'] # --- Force de tendance --- dataframe['adx'] = ta.trend.ADXIndicator( high=dataframe['high'], low=dataframe['low'], close=dataframe['close'], window=14 ).adx() # --- Volume directionnel (On Balance Volume) --- dataframe['obv'] = ta.volume.OnBalanceVolumeIndicator( close=dataframe['close'], volume=dataframe['volume'] ).on_balance_volume() # --- Volatilité récente (écart-type des rendements) --- dataframe['vol_24'] = dataframe['percent'].rolling(24).std() # Compter les baisses / hausses consécutives self.calculateDownAndUp(dataframe, limit=0.0001) # df : ton dataframe OHLCV + indicateurs existants # Assurez-vous que les colonnes suivantes existent : # 'max_rsi_12', 'roc_24', 'bb_percent_1h' # --- Filtrage des NaN initiaux --- # dataframe = dataframe.dropna() dataframe['rsi_slope'] = dataframe['rsi'].diff(3) / 3 # vitesse moyenne du RSI dataframe['adx_change'] = dataframe['adx'] - dataframe['adx'].shift(12) # évolution de la tendance dataframe['volatility_ratio'] = dataframe['atr_norm'] / dataframe['bb_width'] dataframe["rsi_diff"] = dataframe["rsi"] - dataframe["rsi"].shift(3) dataframe["slope_ratio"] = dataframe["sma5_deriv1"] / (dataframe["sma60_deriv1"] + 1e-9) dataframe["divergence"] = (dataframe["rsi_deriv1"] * dataframe["sma5_deriv1"]) < 0 ########################### dataframe['volume_sma_deriv'] = dataframe['volume'] * dataframe['sma5_deriv1'] / (dataframe['volume'].rolling(5).mean()) self.calculeDerivees(dataframe, 'volume', timeframe=timeframe, ema_period=12) self.setTrends(dataframe) return dataframe def feature_auc_scores(self, X, y): aucs = {} for col in X.columns: try: aucs[col] = roc_auc_score(y, X[col].ffill().fillna(0)) except Exception: aucs[col] = np.nan return pd.Series(aucs).sort_values(ascending=False) def macd_tendance_int(self, dataframe: pd.DataFrame, macd_col='macd', signal_col='macdsignal', hist_col='macdhist', eps=0.0) -> pd.Series: """ Renvoie la tendance MACD sous forme d'entiers. 2 : Haussier 1 : Ralentissement hausse 0 : Neutre -1 : Ralentissement baisse -2 : Baissier """ # | Nom | Formule / définition | Signification | # | ---------------------------- | ------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | # | **MACD** (`macd`) | `EMA_fast - EMA_slow` (ex : 12-26 périodes) | Montre l’écart entre la moyenne courte et la moyenne longue.
- Positive → tendance haussière
- Négative → tendance baissière | # | **Signal** (`macdsignal`) | `EMA_9(MACD)` | Sert de ligne de **signal de déclenchement**.
- Croisement du MACD au-dessus → signal d’achat
- Croisement du MACD en dessous → signal de vente | # | **Histogramme** (`macdhist`) | `MACD - Signal` | Montre la **force et l’accélération** de la tendance.
- Positif et croissant → tendance haussière qui s’accélère
- Positif mais décroissant → ralentissement de la hausse
- Négatif et décroissant → baisse qui s’accélère
- Négatif mais croissant → ralentissement de la baisse | # | Situation | MACD | Signal | Hist | Interprétation | # | -------------------------- | ---------- | --------- | -------- | ------------------------------------------ | # | MACD > 0, Hist croissant | au-dessus | croissant | Haussier | Momentum fort → tendance haussière | # | MACD > 0, Hist décroissant | au-dessus | en baisse | Momentum | La hausse ralentit, prudence | # | MACD < 0, Hist décroissant | en dessous | en baisse | Baissier | Momentum fort → tendance baissière | # | MACD < 0, Hist croissant | en dessous | en hausse | Rebond ? | La baisse ralentit → possible retournement | # Créer une série de 0 par défaut tendance = pd.Series(0, index=dataframe.index) # Cas MACD > signal mask_up = dataframe[macd_col] > dataframe[signal_col] + eps mask_up_hist_pos = mask_up & (dataframe[hist_col] > 0) mask_up_hist_neg = mask_up & (dataframe[hist_col] <= 0) tendance[mask_up_hist_pos] = 2 # Haussier tendance[mask_up_hist_neg] = 1 # Ralentissement hausse # Cas MACD < signal mask_down = dataframe[macd_col] < dataframe[signal_col] - eps mask_down_hist_neg = mask_down & (dataframe[hist_col] < 0) mask_down_hist_pos = mask_down & (dataframe[hist_col] >= 0) tendance[mask_down_hist_neg] = -2 # Baissier tendance[mask_down_hist_pos] = -1 # Ralentissement baisse # Les NaN deviennent neutre tendance[dataframe[[macd_col, signal_col, hist_col]].isna().any(axis=1)] = 0 return tendance def calculateDownAndUp(self, dataframe, limit=0.0001): dataframe['down'] = dataframe['hapercent'] <= limit dataframe['up'] = dataframe['hapercent'] >= limit dataframe['down_count'] = - dataframe['down'].astype(int) * ( dataframe['down'].groupby((dataframe['down'] != dataframe['down'].shift()).cumsum()).cumcount() + 1) dataframe['up_count'] = dataframe['up'].astype(int) * ( dataframe['up'].groupby((dataframe['up'] != dataframe['up'].shift()).cumsum()).cumcount() + 1) # Créer une colonne vide dataframe['down_pct'] = self.calculateUpDownPct(dataframe, 'down_count') dataframe['up_pct'] = self.calculateUpDownPct(dataframe, 'up_count') def calculateDerivation(self, dataframe, window=12, suffixe='', timeframe='5m'): dataframe[f"mid_smooth{suffixe}"] = dataframe['mid'].rolling(window).mean() dataframe = self.calculeDerivees(dataframe, f"mid_smooth{suffixe}", timeframe=timeframe, ema_period=window) return dataframe def calculeDerivees( self, dataframe: pd.DataFrame, name: str, suffixe: str = '', window: int = 100, coef: float = 0.15, ema_period: int = 10, verbose: bool = True, timeframe: str = '5m' ) -> pd.DataFrame: """ Calcule deriv1/deriv2 (relative simple), applique EMA, calcule tendency avec epsilon adaptatif basé sur rolling percentiles. """ d1_col = f"{name}{suffixe}_deriv1" d2_col = f"{name}{suffixe}_deriv2" factor1 = 100 * (ema_period / 5) factor2 = 10 * (ema_period / 5) dataframe[f"{name}{suffixe}_inv"] = (dataframe[f"{name}{suffixe}"].shift(2) >= dataframe[f"{name}{suffixe}"].shift(1)) \ & (dataframe[f"{name}{suffixe}"].shift(1) <= dataframe[f"{name}{suffixe}"]) # --- Distance à la moyenne mobile --- dataframe[f"{name}{suffixe}_dist"] = (dataframe['close'] - dataframe[f"{name}{suffixe}"]) / dataframe[f"{name}{suffixe}"] # dérivée relative simple dataframe[d1_col] = 1000 * (dataframe[name] - dataframe[name].shift(1)) / dataframe[name].shift(1) dataframe[d2_col] = dataframe[d1_col] - dataframe[d1_col].shift(1) return dataframe def getOpenTrades(self): # if len(self.trades) == 0: self.trades = Trade.get_open_trades() return self.trades def calculateProbabilite2Index(self, df, futur_cols, indic_1, indic_2): # # Définition des tranches pour les dérivées # bins_deriv = [-np.inf, -0.05, -0.01, 0.01, 0.05, np.inf] # labels = ['forte baisse', 'légère baisse', 'neutre', 'légère hausse', 'forte hausse'] # # # Ajout des colonnes bin (catégorisation) # df[f"{indic_1}_bin"] = pd.cut(df['mid_smooth_1h_deriv1'], bins=bins_deriv, labels=labels) # df[f"{indic_2}_bin"] = pd.cut(df['mid_smooth_deriv1_1d'], bins=bins_deriv, labels=labels) # # # Colonnes de prix futur à analyser # futur_cols = ['futur_percent_1h', 'futur_percent_2h', 'futur_percent_3h', 'futur_percent_4h', 'futur_percent_5h'] # # # Calcul des moyennes et des effectifs # grouped = df.groupby([f"{indic_2}_bin", f"{indic_1}_bin"])[futur_cols].agg(['mean', 'count']) # # pd.set_option('display.width', 200) # largeur max affichage # pd.set_option('display.max_columns', None) pd.set_option('display.max_columns', None) pd.set_option('display.width', 300) # largeur max affichage # nettoyage # series = df[f"{indic_2}"].dropna() # unique_vals = df[f"{indic_2}"].nunique() # print(unique_vals) # print(df[f"{indic_2}"]) n = len(self.labels) df[f"{indic_1}_bin"], bins_1h = pd.qcut(df[f"{indic_1}"], q=n, labels=self.labels, retbins=True, duplicates='drop') df[f"{indic_2}_bin"], bins_1d = pd.qcut(df[f"{indic_2}"], q=n, labels=self.labels, retbins=True, duplicates='drop') # Affichage formaté pour code Python print(f"Bornes des quantiles pour {indic_1} : [{', '.join([f'{b:.4f}' for b in bins_1h])}]") print(f"Bornes des quantiles pour {indic_2} : [{', '.join([f'{b:.4f}' for b in bins_1d])}]") # Agrégation grouped = df.groupby([f"{indic_2}_bin", f"{indic_1}_bin"], observed=True)[futur_cols].agg(['mean', 'count']) # Affichage with pd.option_context('display.max_rows', None, 'display.max_columns', None): print(grouped.round(4)) # Ajout des probabilités de hausse for col in futur_cols: df[f"{col}_is_up"] = df[col] > 0 # Calcul de la proba de hausse proba_up = df.groupby([f"{indic_2}_bin", f"{indic_1}_bin"], observed=True)[f"{col}_is_up"].mean().unstack() print(f"\nProbabilité de hausse pour {col} (en %):") with pd.option_context('display.max_rows', None, 'display.max_columns', None): print((proba_up * 100).round(1)) # Affichage formaté des valeurs comme tableau Python with pd.option_context('display.max_rows', None, 'display.max_columns', None): df_formatted = (proba_up * 100).round(1) print("data = {") for index, row in df_formatted.iterrows(): row_values = ", ".join([f"{val:.1f}" for val in row]) print(f"'{index}': [{row_values}], ") print("}") data = {} for index, row in df_formatted.iterrows(): # on convertit proprement avec arrondi comme dans ton print, mais en données réelles data[index] = [ None if (isinstance(val, float) and math.isnan(val)) else val for val in row ] # Niveaux unicode pour les barres verticales (style sparkline) # spark_chars = "▁▂▃▄▅▆▇█" # print(data.values()) # # Collecte globale min/max # all_values = [] # for vals in data.values(): # all_values.extend(v for v in vals if not (isinstance(v, float) and math.isnan(v))) # # global_min = min(all_values) if all_values else 0 # global_max = max(all_values) if all_values else 1 # global_span = (global_max - global_min) if global_max != global_min else 1 # # def sparkline_global(values): # if all(isinstance(v, float) and math.isnan(v) for v in values): # return "(no data)" # out = "" # for v in values: # if isinstance(v, float) and math.isnan(v): # out += " " # else: # idx = int((v - global_min) / global_span * (len(spark_chars) - 1)) # out += spark_chars[idx] # return out # # for key, values in data.items(): # print(f"{key:>3} : {sparkline_global(values)}") # Palette ANSI 256 couleurs pour heatmap def get_ansi_color(val): """ Échelle fixe 0→100 : 0-20 : bleu (21) 20-40 : cyan (51) 40-60 : vert/jaune (46 / 226) 60-80 : orange (208) 80-100 : rouge (196) """ if val is None: return "" if val < 0: val = 0 elif val > 100: val = 100 if val <= 20: code = 21 elif val <= 40: code = 51 elif val <= 60: code = 226 elif val <= 80: code = 208 else: code = 196 return f"\033[38;5;{code}m" RESET = "\033[0m" # Affichage columns = ['B3', 'B2', 'B1', 'N0', 'H1', 'H2', 'H3'] header = " " + " ".join([f"{col:>6}" for col in columns]) print(header) print("-" * len(header)) for key, values in data.items(): line = f"{key:>3} |" for v in values: if v is None: line += f" {' '} " # vide pour NaN / None else: color = get_ansi_color(v) line += f" {color}{v:5.1f}{RESET} " print(line) def populate_buy_trend(self, dataframe: DataFrame, metadata: dict) -> DataFrame: dataframe.loc[ ( ( ( (dataframe['mid_future_pred_cons'].shift(2) > dataframe['mid_future_pred_cons'].shift(1)) & (dataframe['mid_future_pred_cons'].shift(1) < dataframe['mid_future_pred_cons']) & (dataframe['percent12'] < -0.0005) ) | ( (dataframe['mid_future_pred_cons'] < dataframe['min12']) ) ) & ( ((dataframe['mid_smooth_12_deriv1'] > 0) | (dataframe['mid_smooth_5_deriv1'] > 0)) ) ), ['enter_long', 'enter_tag']] = (1, f"future") dataframe['test'] = np.where(dataframe['enter_long'] == 1, dataframe['close'] * 1.01, np.nan) if self.dp.runmode.value in ('backtest'): dataframe.to_feather(f"user_data/backtest_results/{metadata['pair'].replace('/', '_')}_df.feather") return dataframe def populate_sell_trend(self, dataframe: DataFrame, metadata: dict) -> DataFrame: # dataframe.loc[ # ( # ( # ( # (dataframe['ml_prob'].shift(2) < dataframe['ml_prob'].shift(1)) # & (dataframe['ml_prob'].shift(1) > dataframe['ml_prob']) # ) # | (dataframe['ml_prob'] < 0) # ) # & (dataframe['hapercent'] < 0) # ), ['exit_long', 'exit_tag']] = (1, f"sma60_future") # dataframe.loc[ # ( # ( # ( # (dataframe['mid_future_pred_cons'].shift(2) < dataframe['mid_future_pred_cons'].shift(1)) # & (dataframe['mid_future_pred_cons'].shift(1) > dataframe['mid_future_pred_cons']) # ) # # | (dataframe['mid_smooth_12_deriv1'] < 0) # ) # & (dataframe['sma60_future_pred_cons'] < dataframe['sma60_future_pred_cons'].shift(1)) # & (dataframe['hapercent'] < 0) # ), ['exit_long', 'exit_tag']] = (1, f"sma60_future") # # dataframe.loc[ # ( # ( # (dataframe['mid_future_pred_cons'].shift(2) < dataframe['mid_future_pred_cons'].shift(1)) # & (dataframe['mid_future_pred_cons'].shift(1) > dataframe['mid_future_pred_cons']) # # ) # # & (dataframe['mid_future_pred_cons'] > dataframe['max12']) # & (dataframe['hapercent'] < 0) # # ), ['exit_long', 'exit_tag']] = (1, f"max12") return dataframe def adjust_trade_position(self, trade: Trade, current_time: datetime, current_rate: float, current_profit: float, min_stake: float, max_stake: float, **kwargs): # ne rien faire si ordre deja en cours if trade.has_open_orders: # print("skip open orders") return None if (self.wallets.get_available_stake_amount() < 10): # or trade.stake_amount >= max_stake: return 0 dataframe, _ = self.dp.get_analyzed_dataframe(trade.pair, self.timeframe) last_candle = dataframe.iloc[-1].squeeze() before_last_candle = dataframe.iloc[-2].squeeze() # prépare les données current_time = current_time.astimezone(timezone.utc) open_date = trade.open_date.astimezone(timezone.utc) dispo = round(self.wallets.get_available_stake_amount()) hours_since_first_buy = (current_time - trade.open_date_utc).seconds / 3600.0 days_since_first_buy = (current_time - trade.open_date_utc).days hours = (current_time - trade.date_last_filled_utc).total_seconds() / 3600.0 count_of_buys = trade.nr_of_successful_entries current_time_utc = current_time.astimezone(timezone.utc) open_date = trade.open_date.astimezone(timezone.utc) days_since_open = (current_time_utc - open_date).days pair = trade.pair profit = trade.calc_profit(current_rate) #round(current_profit * trade.stake_amount, 1) last_lost = self.getLastLost(last_candle, pair) pct_first = 0 total_counts = sum( pair_data['count_of_buys'] for pair_data in self.pairs.values() if not self.getShortName(pair) == 'BTC') if self.pairs[pair]['first_buy']: pct_first = self.getPctFirstBuy(pair, last_candle) pct = self.pct.value if count_of_buys == 1: pct_max = current_profit else: if self.pairs[trade.pair]['last_buy']: pct_max = self.getPctLastBuy(pair, last_candle) else: pct_max = - pct if (self.getShortName(pair) == 'BTC') or count_of_buys <= 2: lim = - pct - (count_of_buys * self.pct_inc.value) else: pct = 0.05 lim = - pct - (count_of_buys * 0.0025) if (len(dataframe) < 1): # print("skip dataframe") return None if not self.should_enter_trade(pair, last_candle, current_time): return None condition = (last_candle['enter_long'] and last_candle['stop_buying_1h'] == False and last_candle['hapercent'] > 0) # and last_candle['sma60_deriv1'] > 0 # or last_candle['enter_tag'] == 'pct3' \ # or last_candle['enter_tag'] == 'pct3_1h' # if (self.getShortName(pair) != 'BTC' and count_of_buys > 3): # condition = before_last_candle_24['mid_smooth_3_1h'] > before_last_candle_12['mid_smooth_3_1h'] and before_last_candle_12['mid_smooth_3_1h'] < last_candle['mid_smooth_3_1h'] #and last_candle['mid_smooth_3_deriv1_1h'] < -1.5 limit_buy = 40 if (count_of_buys < limit_buy) and condition and (pct_max < lim): try: if self.pairs[pair]['has_gain'] and profit > 0: self.pairs[pair]['force_sell'] = True return None max_amount = self.config.get('stake_amount') * 2.5 stake_amount = min(min(max_amount, self.wallets.get_available_stake_amount()), self.adjust_stake_amount(pair, last_candle) * abs(last_lost / self.mise_factor_buy.value)) if stake_amount > 0: trade_type = "Loss " + (last_candle['enter_tag'] if last_candle['enter_long'] == 1 else '') self.pairs[trade.pair]['count_of_buys'] += 1 self.pairs[pair]['total_amount'] += stake_amount self.log_trade( last_candle=last_candle, date=current_time, action="🟧 Loss -", dispo=dispo, pair=trade.pair, rate=current_rate, trade_type=trade_type, profit=round(profit, 1), buys=trade.nr_of_successful_entries + 1, stake=round(stake_amount, 2) ) self.pairs[trade.pair]['last_buy'] = current_rate self.pairs[trade.pair]['max_touch'] = last_candle['close'] self.pairs[trade.pair]['last_candle'] = last_candle # df = pd.DataFrame.from_dict(self.pairs, orient='index') # colonnes_a_exclure = ['last_candle', 'stop', # 'trade_info', 'last_date', 'expected_profit', 'last_count_of_buys', 'base_stake_amount', 'stop_buy'] # df_filtered = df[df['count_of_buys'] > 0].drop(columns=colonnes_a_exclure) # # df_filtered = df_filtered["first_buy", "last_max", "max_touch", "last_sell","last_buy", 'count_of_buys', 'current_profit'] # # print(df_filtered) return stake_amount return None except Exception as exception: print(exception) return None if (profit > self.pairs[pair]['previous_profit'] and profit > self.pairs[pair]['expected_profit'] and hours > 6 # and last_candle['sma60_deriv1'] > 0 and last_candle['max_rsi_12_1h'] < 75 # and last_candle['rsi_1d'] < 58 # and last_candle['stop_buying'] == False # and last_candle['mid_smooth_5_deriv1_1d'] > 0 and self.wallets.get_available_stake_amount() > 0 ): try: self.pairs[pair]['previous_profit'] = profit stake_amount = min(self.wallets.get_available_stake_amount(), self.pairs[pair]['first_amount']) if stake_amount > 0: self.pairs[pair]['has_gain'] += 1 trade_type = 'Gain +' + (last_candle['enter_tag'] if last_candle['enter_long'] == 1 else '') self.pairs[trade.pair]['count_of_buys'] += 1 self.pairs[pair]['total_amount'] += stake_amount self.log_trade( last_candle=last_candle, date=current_time, action="🟡 Gain +", dispo=dispo, pair=trade.pair, rate=current_rate, trade_type=str(round(pct_max, 4)), profit=round(profit, 1), buys=trade.nr_of_successful_entries + 1, stake=round(stake_amount, 2) ) self.pairs[trade.pair]['last_buy'] = current_rate self.pairs[trade.pair]['max_touch'] = last_candle['close'] self.pairs[trade.pair]['last_candle'] = last_candle return stake_amount return None except Exception as exception: print(exception) return None return None def getPctFirstBuy(self, pair, last_candle): return round((last_candle['close'] - self.pairs[pair]['first_buy']) / self.pairs[pair]['first_buy'], 3) def getPctLastBuy(self, pair, last_candle): return round((last_candle['close'] - self.pairs[pair]['last_buy']) / self.pairs[pair]['last_buy'], 4) def adjust_stake_amount(self, pair: str, last_candle: DataFrame): # Calculer le minimum des 14 derniers jours nb_pairs = len(self.dp.current_whitelist()) base_stake_amount = self.config.get('stake_amount') / (self.mises.value) # * nb_pairs) # Montant de base configuré # factors = [1, 1.2, 1.3, 1.4] if self.pairs[pair]['count_of_buys'] == 0: factor = 1 #65 / min(65, last_candle['rsi_1d']) if last_candle['open'] < last_candle['sma5_1h'] and last_candle['mid_smooth_12_deriv1'] > 0: factor = 2 adjusted_stake_amount = max(base_stake_amount / 5, base_stake_amount * factor) else: adjusted_stake_amount = self.pairs[pair]['first_amount'] if self.pairs[pair]['count_of_buys'] == 0: self.pairs[pair]['first_amount'] = adjusted_stake_amount return adjusted_stake_amount def expectedProfit(self, pair: str, last_candle: DataFrame): lim = 0.01 pct = 0.002 if (self.getShortName(pair) == 'BTC'): lim = 0.005 pct = 0.001 pct_to_max = lim + pct * self.pairs[pair]['count_of_buys'] expected_profit = lim * self.pairs[pair]['total_amount'] # min(3 * lim, max(lim, pct_to_max)) # 0.004 + 0.002 * self.pairs[pair]['count_of_buys'] #min(0.01, first_max) self.pairs[pair]['expected_profit'] = expected_profit return expected_profit def calculateUpDownPct(self, dataframe, key): down_pct_values = np.full(len(dataframe), np.nan) # Remplir la colonne avec les bons calculs for i in range(len(dataframe)): shift_value = abs(int(dataframe[key].iloc[i])) # Récupérer le shift actuel if i - shift_value > 1: # Vérifier que le shift ne dépasse pas l'index down_pct_values[i] = 100 * (dataframe['close'].iloc[i] - dataframe['close'].iloc[i - shift_value]) / \ dataframe['close'].iloc[i - shift_value] return down_pct_values @property def protections(self): return [ { "method": "CooldownPeriod", "stop_duration_candles": 12 } # { # "method": "MaxDrawdown", # "lookback_period_candles": self.lookback.value, # "trade_limit": self.trade_limit.value, # "stop_duration_candles": self.protection_stop.value, # "max_allowed_drawdown": self.protection_max_allowed_dd.value, # "only_per_pair": False # }, # { # "method": "StoplossGuard", # "lookback_period_candles": 24, # "trade_limit": 4, # "stop_duration_candles": self.protection_stoploss_stop.value, # "only_per_pair": False # }, # { # "method": "StoplossGuard", # "lookback_period_candles": 24, # "trade_limit": 4, # "stop_duration_candles": 2, # "only_per_pair": False # }, # { # "method": "LowProfitPairs", # "lookback_period_candles": 6, # "trade_limit": 2, # "stop_duration_candles": 60, # "required_profit": 0.02 # }, # { # "method": "LowProfitPairs", # "lookback_period_candles": 24, # "trade_limit": 4, # "stop_duration_candles": 2, # "required_profit": 0.01 # } ] def get_stake_from_drawdown(self, pct: float, base_stake: float = 100.0, step: float = 0.04, growth: float = 1.15, max_stake: float = 1000.0) -> float: """ Calcule la mise à allouer en fonction du drawdown. :param pct: Drawdown en pourcentage (ex: -0.12 pour -12%) :param base_stake: Mise de base (niveau 0) :param step: Espacement entre paliers (ex: tous les -4%) :param growth: Facteur de croissance par palier (ex: 1.15 pour +15%) :param max_stake: Mise maximale à ne pas dépasser :return: Montant à miser """ if pct >= 0: return base_stake level = int(abs(pct) / step) stake = base_stake * (growth ** level) return min(stake, max_stake) def polynomial_forecast(self, series: pd.Series, window: int = 20, degree: int = 2, steps=[12, 24, 36]): """ Calcule une régression polynomiale sur les `window` dernières valeurs de la série, puis prédit les `n_future` prochaines valeurs. :param series: Série pandas (ex: dataframe['close']) :param window: Nombre de valeurs récentes utilisées pour ajuster le polynôme :param degree: Degré du polynôme (ex: 2 pour quadratique) :param n_future: Nombre de valeurs futures à prédire :return: tuple (poly_function, x_vals, y_pred), où y_pred contient les prédictions futures """ if len(series) < window: raise ValueError("La série est trop courte pour la fenêtre spécifiée.") recent_y = series.iloc[-window:].values x = np.arange(window) coeffs = np.polyfit(x, recent_y, degree) poly = np.poly1d(coeffs) x_future = np.arange(window, window + len(steps)) y_future = poly(x_future) # Affichage de la fonction # print("Fonction polynomiale trouvée :") # print(poly) current = series.iloc[-1] count = 0 for future_step in steps: # range(1, n_future + 1) future_x = window - 1 + future_step prediction = poly(future_x) # series.loc[series.index[future_x], f'poly_pred_t+{future_step}'] = prediction # ➕ Afficher les prédictions # print(f"{current} → t+{future_step}: x={future_x}, y={prediction:.2f}") if prediction > 0: # current: count += 1 return poly, x_future, y_future, count def should_enter_trade(self, pair: str, last_candle, current_time) -> bool: limit = 3 # if self.pairs[pair]['stop'] and last_candle['max_rsi_12_1h'] <= 60 and last_candle['trend_class_1h'] == -1: # dispo = round(self.wallets.get_available_stake_amount()) # self.pairs[pair]['stop'] = False # self.log_trade( # last_candle=last_candle, # date=current_time, # action="🟢RESTART", # dispo=dispo, # pair=pair, # rate=last_candle['close'], # trade_type='', # profit=0, # buys=self.pairs[pair]['count_of_buys'], # stake=0 # ) # 🟢 Dérivée 1 > 0 et dérivée 2 > 0: tendance haussière qui s’accélère. # 🟡 Dérivée 1 > 0 et dérivée 2 < 0: tendance haussière qui ralentit → essoufflement potentiel. # 🔴 Dérivée 1 < 0 et dérivée 2 < 0: tendance baissière qui s’accélère. # 🟠 Dérivée 1 < 0 et dérivée 2 > 0: tendance baissière qui ralentit → possible bottom. # if not pair.startswith('BTC'): dispo = round(self.wallets.get_available_stake_amount()) # if self.pairs[pair]['stop'] \ # and last_candle[f"{self.indic_1d_p.value}_deriv1_1h"] >= self.indic_deriv1_1d_p_start.value \ # and last_candle[f"{self.indic_1d_p.value}_deriv2_1h"] >= self.indic_deriv2_1d_p_start.value: # self.pairs[pair]['stop'] = False # self.log_trade( # last_candle=last_candle, # date=current_time, # action="🟢RESTART", # dispo=dispo, # pair=pair, # rate=last_candle['close'], # trade_type='', # profit=0, # buys=self.pairs[pair]['count_of_buys'], # stake=0 # ) # else: # if self.pairs[pair]['stop'] == False \ # and last_candle[f"{self.indic_1d_p.value}_deriv1_1h"] <= self.indic_deriv1_1d_p_stop.value \ # and last_candle[f"{self.indic_1d_p.value}_deriv2_1h"] <= self.indic_deriv2_1d_p_stop.value: # self.pairs[pair]['stop'] = True # # if self.pairs[pair]['current_profit'] > 0: # # self.pairs[pair]['force_sell'] = True # self.log_trade( # last_candle=last_candle, # date=current_time, # action="🔴STOP", # dispo=dispo, # pair=pair, # rate=last_candle['close'], # trade_type='', # profit=self.pairs[pair]['current_profit'], # buys=self.pairs[pair]['count_of_buys'], # stake=0 # ) # return False # if self.pairs[pair]['stop']: # return False return True # Filtrer les paires non-BTC non_btc_pairs = [p for p in self.pairs if not p.startswith('BTC')] # Compter les positions actives sur les paires non-BTC max_nb_trades = 0 total_non_btc = 0 max_pair = '' limit_amount = 250 max_amount = 0 for p in non_btc_pairs: max_nb_trades = max(max_nb_trades, self.pairs[p]['count_of_buys']) max_amount = max(max_amount, self.pairs[p]['total_amount']) for p in non_btc_pairs: if (max_nb_trades == self.pairs[p]['count_of_buys'] and max_nb_trades > limit): # if (max_amount == self.pairs[p]['total_amount'] and max_amount > limit_amount): max_pair = p total_non_btc += self.pairs[p]['count_of_buys'] pct_max = self.getPctFirstBuy(pair, last_candle) # self.getPctLastBuy(pair, last_candle) if last_candle['mid_smooth_1h_deriv1'] < -0.02: # and last_candle['mid_smooth_1h_deriv2'] > 0): return False self.should_enter_trade_count = 0 # if max_pair != pair and self.pairs[pair]['total_amount'] > 300: # return False if (max_pair != '') & (self.pairs[pair]['count_of_buys'] >= limit): trade = self.pairs[max_pair]['current_trade'] current_time = current_time.astimezone(timezone.utc) open_date = trade.open_date.astimezone(timezone.utc) current_time_utc = current_time.astimezone(timezone.utc) days_since_open = (current_time_utc - open_date).days pct_max_max = self.getPctFirstBuy(max_pair, last_candle) # print(f"days_since_open {days_since_open} max_pair={max_pair} pair={pair}") return max_pair == pair or pct_max < - 0.25 or ( pct_max_max < - 0.15 and max_pair != pair and days_since_open > 30) else: return True def select_uncorrelated_features(self, df, target, top_n=20, corr_threshold=0.7): """ Sélectionne les features les plus corrélées avec target, tout en supprimant celles trop corrélées entre elles. """ # 1️⃣ Calcul des corrélations absolues avec la cible corr = df.corr(numeric_only=True) corr_target = corr[target].abs().sort_values(ascending=False) # 2️⃣ Prend les N features les plus corrélées avec la cible (hors target) features = corr_target.drop(target).head(top_n).index.tolist() # 3️⃣ Évite les features trop corrélées entre elles selected = [] for feat in features: too_correlated = False for sel in selected: if abs(corr.loc[feat, sel]) > corr_threshold: too_correlated = True break if not too_correlated: selected.append(feat) # 4️⃣ Retourne un DataFrame propre avec les valeurs de corrélation selected_corr = pd.DataFrame({ "feature": selected, "corr_with_target": [corr.loc[f, target] for f in selected] }).sort_values(by="corr_with_target", key=np.abs, ascending=False) return selected_corr def graphFonctionApprise(self, path, X_test, y_test, y_pred): # Exemple : trier les valeurs de X_test et les prédictions x_sorted = np.argsort(X_test.iloc[:, 0]) x = X_test.iloc[:, 0].iloc[x_sorted] y_true = y_test.iloc[x_sorted] y_pred = y_pred[x_sorted] plt.figure(figsize=(12, 6)) plt.plot(x, y_true, label="Réel", color="blue", alpha=0.7) plt.plot(x, y_pred, label="Prédit (LGBM)", color="red", alpha=0.7) plt.title("Fonction apprise par LGBMRegressor") plt.xlabel("Feature principale") plt.ylabel("Valeur prédite") plt.legend() plt.grid(True) out_path = f"{self.path}/lgbm_function.png" plt.savefig(out_path, bbox_inches="tight") plt.close() print(f"Graphique sauvegardé : {out_path}") def graphFonctionAppriseFeature(self, path, X_test, y_test, y_pred): plt.figure(figsize=(14, 8)) colors = sns.color_palette("coolwarm", n_colors=X_test.shape[1]) # Conversion en DataFrame pour manip plus simple df = X_test.copy() df["y_pred"] = y_pred # --- filtrage sur y_pred (ou sur chaque feature si tu veux) mean = df["y_pred"].mean() std = df["y_pred"].std() df = df[(df["y_pred"] >= mean - 2 * std) & (df["y_pred"] <= mean + 2 * std)] # --- tracé for i, col in enumerate(X_test.columns): plt.plot(df[col], df["y_pred"], '.', color=colors[i], alpha=0.4, label=col) plt.title("Fonction apprise par LGBMRegressor (filtrée à ±2σ)") plt.xlabel("Valeur feature") plt.ylabel("Valeur prédite") plt.legend(loc="right") plt.grid(True) out_path = f"{self.path}/lgbm_features.png" plt.savefig(out_path, bbox_inches="tight") plt.close() print(f"Graphique sauvegardé : {out_path}") def optuna(self, path, X_train, X_test, y_train, y_test): # Suppose que X_train, y_train sont déjà définis # ou sinon : # X_train, X_valid, y_train, y_valid = train_test_split(X, y, test_size=0.2, random_state=42) print("Description") print(X_train.describe().T.sort_values("std")) def objective(trial): params = { 'objective': 'regression', 'metric': 'rmse', 'n_estimators': trial.suggest_int('n_estimators', 100, 1000), 'learning_rate': trial.suggest_float('learning_rate', 0.005, 0.2, log=True), 'max_depth': trial.suggest_int('max_depth', 3, 15), 'num_leaves': trial.suggest_int('num_leaves', 20, 300), 'subsample': trial.suggest_float('subsample', 0.5, 1.0), 'colsample_bytree': trial.suggest_float('colsample_bytree', 0.5, 1.0), 'reg_alpha': trial.suggest_float('reg_alpha', 0.0, 10.0), 'reg_lambda': trial.suggest_float('reg_lambda', 0.0, 10.0), 'random_state': 42, } model = LGBMRegressor(**params) model.fit(X_train, y_train) # On peut aussi valider sur un split interne preds = model.predict(X_test) rmse = np.sqrt(mean_squared_error(y_test, preds)) return rmse # Crée une étude Optuna study = optuna.create_study(direction="minimize") # on veut minimiser l'erreur study.optimize(objective, n_trials=50, show_progress_bar=True) # 🔹 Afficher les meilleurs résultats print("✅ Meilleurs hyperparamètres trouvés :") print(study.best_params) print(f"Meilleur RMSE : {study.best_value:.4f}") # 🔹 Sauvegarder les résultats optuna_path = f"{self.path}/optuna_lgbm_results.txt" with open(optuna_path, "w") as f: f.write(f"Best params:\n{study.best_params}\n") f.write(f"Best RMSE: {study.best_value:.4f}\n") print(f"Résultats sauvegardés dans : {optuna_path}") # 🔹 Créer le modèle final avec les meilleurs paramètres print("🚀 Entraînement du modèle LightGBM...") # -- Appliquer le filtrage -- X_train_filtered = self.filter_features(X_train, y_train) best_model = LGBMRegressor(**study.best_params) best_model.fit(X_train_filtered, y_train) # fig1 = vis.plot_optimization_history(study) # fig1.write_image("/home/souti/freqtrade/user_data/plots/optuna_history.png") # # fig2 = vis.plot_param_importances(study) # fig2.write_image("/home/souti/freqtrade/user_data/plots/optuna_importance.png") return best_model, X_train_filtered def filter_features(self, X: pd.DataFrame, y: pd.Series, corr_threshold: float = 0.95): """Filtre les colonnes peu utiles ou redondantes""" print("🔍 Filtrage automatique des features...") # 1️⃣ Supprimer les colonnes constantes vt = VarianceThreshold(threshold=1e-5) X_var = pd.DataFrame(vt.fit_transform(X), columns=X.columns[vt.get_support()]) print(f" - {len(X.columns) - X_var.shape[1]} colonnes supprimées (variance faible)") # 2️⃣ Supprimer les colonnes très corrélées entre elles corr = X_var.corr().abs() upper = corr.where(np.triu(np.ones(corr.shape), k=1).astype(bool)) drop_cols = [column for column in upper.columns if any(upper[column] > corr_threshold)] X_corr = X_var.drop(columns=drop_cols, errors='ignore') print(f" - {len(drop_cols)} colonnes supprimées (corrélation > {corr_threshold})") # 3️⃣ Facultatif : supprimer les colonnes entièrement NaN X_clean = X_corr.dropna(axis=1, how='all') print(f"✅ {X_clean.shape[1]} colonnes conservées après filtrage.\n") return X_clean def setTrends(self, dataframe: DataFrame): SMOOTH_WIN=10 df = dataframe.copy() # # --- charger les données --- # df['timestamp'] = pd.to_datetime(df['timestamp'], errors='coerce') # --- calcul SMA14 --- # df['sma'] = talib.SMA(df, timeperiod=20) # ta.trend.sma_indicator(df['close'], 14) # --- pente brute --- df['slope'] = df['sma12'].diff() # --- lissage EMA --- df['slope_smooth'] = df['slope'].ewm(span=SMOOTH_WIN, adjust=False).mean() # df["slope_smooth"] = savgol_filter(df["slope_smooth"], window_length=21, polyorder=3) # --- normalisation relative --- df['slope_norm'] = 10000 * df['slope_smooth'] / df['close'] # df['slope_norm'].fillna(0, inplace=True) df['slope_norm'] = df['slope_norm'].fillna(0) dataframe['slope_norm'] = df['slope_norm'] try: from lightgbm import LGBMRegressor _HAS_LGBM = True except Exception: _HAS_LGBM = False def make_model(self, model_type="linear", degree=2, random_state=0): model_type = model_type.lower() if model_type == "linear": return LinearRegression() if model_type == "poly": return make_pipeline(StandardScaler(), PolynomialFeatures(degree=degree, include_bias=False), LinearRegression()) if model_type == "svr": return make_pipeline(StandardScaler(), SVR(kernel="rbf", C=1.0, epsilon=0.1)) if model_type == "rf": return RandomForestRegressor(n_estimators=100, random_state=random_state, n_jobs=1) if model_type == "lgbm": if not _HAS_LGBM: raise RuntimeError("lightgbm n'est pas installé") return LGBMRegressor(n_estimators=100, random_state=random_state) raise ValueError(f"model_type inconnu: {model_type}") def calculateRegressionNew(self, df, indic, lookback=20, future_steps=5, model_type="linear"): df = df.copy() pred_col = f"{indic}_future_pred_cons" df[pred_col] = np.nan X_idx = np.arange(lookback).reshape(-1, 1) values = df[indic].values n = len(values) model = LinearRegression() for i in range(lookback, n - future_steps): window = values[i - lookback:i] # cible = vraie valeur future y_target = values[i + future_steps] if np.isnan(window).any() or np.isnan(y_target): continue # entraînement model.fit(X_idx, window) # prédiction de la valeur future future_x = np.array([[lookback + future_steps - 1]]) pred_future = model.predict(future_x)[0] # la prédiction concerne i + future_steps df.iloc[i + future_steps, df.columns.get_loc(pred_col)] = pred_future return df # ========================================================== # NOUVELLE VERSION : calcule AUSSI les dernières valeurs ! # ========================================================== def calculateRegression( self, df, indic, lookback=30, future_steps=5, model_type="linear", degree=2, weight_mode="exp", weight_strength=2, clip_k=2.0, blend_alpha=0.7, ): values = df[indic].values.astype(float) n = len(values) colname = f"{indic}_future_pred_cons" df[colname] = np.nan # pré-calcul des fenêtres windows = np.lib.stride_tricks.sliding_window_view(values, lookback) # windows[k] = valeurs de [k .. k+lookback-1] # indices valides d’entraînement trainable_end = n - future_steps # créer une fois le modèle model = self.make_model(model_type=model_type, degree=degree) # ================ # BOUCLE TRAINING # ================ for i in range(lookback, trainable_end): window = values[i - lookback:i] if np.isnan(window).any(): continue # delta future réelle y_target = values[i + future_steps] - values[i] # features = positions dans la fenêtre : 0..lookback-1 X_window = np.arange(lookback).reshape(-1, 1) # sample weights if weight_mode == "exp": weights = np.linspace(0.1, 1, lookback) ** weight_strength else: weights = None # entraînement try: model.fit(X_window, window, sample_weight=weights) except Exception: model.fit(X_window, window) # prédiction de la valeur future (position lookback+future_steps-1) y_pred_value = model.predict( np.array([[lookback + future_steps - 1]]) )[0] pred_delta = y_pred_value - values[i] # clipping par volatilité locale local_std = np.std(window) max_change = clip_k * (local_std if local_std > 0 else 1e-9) pred_delta = np.clip(pred_delta, -max_change, max_change) # blend final_pred_value = ( blend_alpha * (values[i] + pred_delta) + (1 - blend_alpha) * values[i] ) df.iloc[i, df.columns.get_loc(colname)] = final_pred_value # ========================================================== # 🔥 CALCUL DES DERNIÈRES VALEURS MANQUANTES 🔥 # ========================================================== # Il reste les indices : [n - future_steps … n - 1] for i in range(trainable_end, n): # fenêtre glissante de fin if i - lookback < 0: continue window = values[i - lookback:i] if np.isnan(window).any(): continue # features X_window = np.arange(lookback).reshape(-1, 1) try: model.fit(X_window, window) except: continue # prédiction d’une continuation locale : future_steps = 1 en fin y_pred_value = model.predict(np.array([[lookback]]))[0] pred_delta = y_pred_value - values[i - 1] final_pred_value = ( blend_alpha * (values[i - 1] + pred_delta) + (1 - blend_alpha) * values[i - 1] ) df.iloc[i, df.columns.get_loc(colname)] = final_pred_value return df # def calculateRegression(self, # df, # indic, # lookback=30, # future_steps=5, # model_type="linear", # degree=2, # random_state=0, # weight_mode="exp", # "exp", "linear" ou None # weight_strength=0.2, # plus c’est grand, plus les dernières bougies comptent # ): # """ # Ajoute une régression glissante qui prévoit la valeur future à horizon 'future_steps', # avec pondération des dernières valeurs si weight_mode != None. # """ # df = df.copy() # colname = f"{indic}_future_pred_{model_type}" # df[colname] = np.nan # # values = df[indic].values # n = len(values) # X_window = np.arange(lookback).reshape(-1, 1) # # # génération du schéma de pondération # if weight_mode == "exp": # # exponentiel → les derniers points pèsent beaucoup plus # weights = np.exp(np.linspace(-weight_strength, weight_strength, lookback)) # elif weight_mode == "linear": # # poids linéaire → 1..lookback # weights = np.linspace(0.5, 1.0, lookback) # else: # weights = np.ones(lookback) # # for i in range(lookback, n - future_steps): # y_window = values[i - lookback:i] # if np.isnan(y_window).any(): # continue # # model = self.make_model(model_type=model_type, degree=degree, random_state=random_state) # # try: # model.fit(X_window, y_window, sample_weight=weights) # except TypeError: # # certains modèles (RF) ne supportent pas sample_weight dans ce contexte # model.fit(X_window, y_window) # except Exception: # continue # # X_pred = np.array([[lookback + future_steps - 1]]) # try: # pred = model.predict(X_pred)[0] # except Exception: # continue # # df.iloc[i, df.columns.get_loc(colname)] = pred # # return df # def calculateRegression(self, df, indic, lookback=30, future_steps=5): # """ # Ajoute un indicateur {indic}_future_pred qui contient, # pour chaque bougie n, la valeur attendue à n + future_steps # selon une régression linéaire sur les lookback dernières bougies. # """ # df = df.copy() # df[f"{indic}_future_pred"] = np.nan # # values = df[indic].values # n = len(values) # # model = LinearRegression() # # for i in range(lookback, n - future_steps): # # Fenêtre d’apprentissage # X = np.arange(lookback).reshape(-1, 1) # y = values[i - lookback:i] # # model.fit(X, y) # # # Prédiction future # next_X = np.array([[lookback + future_steps - 1]]) # future_pred = model.predict(next_X)[0] # # # On insère la prédiction à la position actuelle (n) # df.iloc[i, df.columns.get_loc(f"{indic}_future_pred")] = future_pred # # return df def add_future_quantiles(self, dataframe, indic, lookback=30, future_steps=5, quantiles=[0.1, 0.5, 0.9]): working_columns = self.listUsableColumns(dataframe) df = dataframe[self.model_indicators].copy() n = len(df) target = self.indicator_target + "_future" df[target] = dataframe[self.indicator_target].shift(-24) # > df['sma24'] * 1.003).astype(int) df[target] = df[target].fillna(0) #.astype(int) # Créer les colonnes pour chaque quantile for q in quantiles: df[f"{indic}_future_q{int(q * 100)}"] = np.nan # Préparer toutes les fenêtres X X = np.array([df[indic].iloc[i - lookback:i].values for i in range(lookback, n - future_steps)]) y_idx = np.arange(lookback, n - future_steps) + future_steps # index des valeurs futures # Imputer les NaN imputer = SimpleImputer(strategy='median') X_imputed = imputer.fit_transform(X) # Pour chaque quantile, créer un modèle et prédire for q in quantiles: model = HistGradientBoostingRegressor(loss='quantile', quantile=q, max_iter=100) # Entrainer chaque ligne X_imputed à prédire la dernière valeur de la fenêtre + future_steps # Ici, comme on prédit delta future par fenêtre, on peut utiliser la valeur cible correspondante y = df[indic].iloc[y_idx].values model.fit(X_imputed, y) y_pred = model.predict(X_imputed) # Écrire les prédictions dans le dataframe df.iloc[lookback:n - future_steps, df.columns.get_loc(f"{indic}_future_q{int(q * 100)}")] = y_pred df_plot = df.iloc[lookback:-future_steps] self.plot_future_quantiles_band(df_plot, indic=self.indicator_target, quantiles=[0.1, 0.5, 0.9]) # self.compute_quantile_confidence(df_plot, indic=self.indicator_target, quantiles=[0.1, 0.5, 0.9]) # fig, ax = plt.subplots(figsize=(20, 20)) # for q in quantiles: # plt.plot(stats.index.astype(str), stats[q], marker='o', label=f"Q{int(q * 100)}") # plt.xticks(rotation=45) # plt.xlabel(f"{indic} bins") # plt.ylabel(f"Quantiles") # plt.title(f"Distribution quantile de {indic}") # plt.legend() # plt.grid(True) # plt.tight_layout() # # plt.show() # # --- Sauvegarde --- # output_path = f"{path}/Distribution_quantile.png" # plt.savefig(output_path, bbox_inches="tight", dpi=150) # plt.close(fig) # # target = "future_return" quantiles = [0.1, 0.25, 0.5, 0.75, 0.9] for indicator in working_columns: df["bin"] = pd.qcut(df[indicator], q=20, duplicates="drop") stats = df.groupby("bin")[target].quantile(quantiles).unstack() fig, ax = plt.subplots(figsize=(10, 10)) # plt.figure(figsize=(12, 6)) for q in stats.columns: plt.plot(stats.index.astype(str), stats[q], marker='o', label=f"Q{int(q * 100)}") plt.xticks(rotation=45) plt.xlabel(f"{indicator} bins") plt.ylabel(f"Quantiles of {target}") plt.title(f"Distribution quantile de {target} selon {indicator}") plt.legend() plt.grid(True) plt.tight_layout() # --- Sauvegarde --- output_path = f"{self.path}/Distribution_{indicator}.png" plt.savefig(output_path, bbox_inches="tight", dpi=150) plt.close(fig) # plt.show() return df def plot_future_quantiles_band(self, df, indic, quantiles=[0.1, 0.5, 0.9], lookback=30, future_steps=5): """ df: DataFrame contenant la colonne réelle et les colonnes de quantiles indic: nom de la colonne cible (ex: 'mid') quantiles: liste des quantiles prédits """ # plt.figure(figsize=(16, 6)) fig, ax = plt.subplots(figsize=(96, 30)) # Série réelle plt.plot(df[indic], label=f"{indic} réel", color='black', linewidth=1.2) # Récupérer les colonnes de quantiles cols_q = [f"{indic}_future_q{int(q * 100)}" for q in quantiles] # Vérifier que tous les quantiles existent cols_q = [c for c in cols_q if c in df.columns] if len(cols_q) < 2: print("Au moins deux quantiles sont nécessaires pour afficher les bandes") return # Ordre : q_min, q_median, q_max df_plot = df[cols_q] # Couleur pour la bande color = sns.color_palette("coolwarm", n_colors=1)[0] # Tracer la bande entre min et max quantiles plt.fill_between(df.index, df_plot.iloc[:, 0], # quantile bas (ex: 10%) df_plot.iloc[:, -1], # quantile haut (ex: 90%) color=color, alpha=0.3, label=f"Intervalle {quantiles[0] * 100}-{quantiles[-1] * 100}%") # Tracer la médiane if len(cols_q) >= 3: plt.plot(df_plot.iloc[:, 1], color=color, linestyle='--', linewidth=1, label="Quantile médian") plt.title(f"Prédiction futures valeurs de {indic} avec intervalle de quantiles") plt.xlabel("Index / Bougies") plt.ylabel(indic) plt.legend() plt.grid(True) # plt.show() # --- Sauvegarde --- output_path = f"{self.path}/Prédiction futures valeurs de {indic}.png" plt.savefig(output_path, bbox_inches="tight", dpi=150) plt.close(fig) def compute_quantile_confidence(self, df, indic, quantiles=[0.1, 0.5, 0.9]): """ df: DataFrame contenant les colonnes des quantiles indic: nom de la colonne réelle quantiles: liste des quantiles prédits Retourne une série score [-1,1], positif = au-dessus de la médiane, négatif = en dessous """ # df['quantile_conf'] = compute_quantile_confidence(df_plot, indic='mid') # # # Exemple de signal simple # df['buy_signal'] = df['quantile_conf'] < -0.5 # valeur sous la médiane + bande étroite # df['sell_signal'] = df['quantile_conf'] > 0.5 # valeur au-dessus de la médiane + bande étroite col_low = f"{indic}_future_q{int(quantiles[0] * 100)}" col_med = f"{indic}_future_q{int(quantiles[1] * 100)}" col_high = f"{indic}_future_q{int(quantiles[2] * 100)}" # largeur de bande (incertitude) band_width = df[col_high] - df[col_low] + 1e-9 # éviter division par 0 # distance normalisée à la médiane score = (df[indic] - df[col_med]) / band_width # clipper le score dans [-1,1] pour éviter les valeurs extrêmes score = np.clip(score, -1, 1) # plt.figure(figsize=(16, 6)) fig, ax = plt.subplots(figsize=(16, 6)) plt.plot(df[indic], color='black', label='Valeur réelle') plt.fill_between(df.index, df[f"{indic}_future_q10"], df[f"{indic}_future_q90"], alpha=0.3, color='blue', label='Intervalle 10%-90%') plt.plot(df[f"{indic}_future_q50"], linestyle='--', color='blue', label='Médiane') # Ajouter le score comme couleur de fond plt.scatter(df.index, df[indic], c=df['quantile_conf'], cmap='coolwarm', s=20) plt.colorbar(label='Score de confiance') plt.title("Prédiction + score de confiance quantile") plt.legend() plt.grid(True) # plt.show() # --- Sauvegarde --- output_path = f"{self.path}/Prédiction score confiance de {indic}.png" plt.savefig(output_path, bbox_inches="tight", dpi=150) plt.close(fig) return score # def loadTensorFlow(self, dataframe, metadata, lookback=50, future_steps=1): # self.model = load_model(f"{self.path}/lstm_model.keras", compile=False) # # # features = toutes les colonnes sauf la cible # feature_columns = self.model_indicators #[col for col in dataframe.columns if col != self.indicator_target] # X_values = dataframe[feature_columns].values # # # normalisation avec le même scaler que l'entraînement # scaler_X = MinMaxScaler() # scaler_X.fit(X_values) # ou charger les paramètres si sauvegardés # X_scaled = scaler_X.transform(X_values) # # # création des fenêtres glissantes # X = np.lib.stride_tricks.sliding_window_view(X_scaled, window_shape=(self.lookback, X_scaled.shape[1])) # # np.lib.stride_tricks.sliding_window_view ne supporte pas directement 2D → il vaut mieux utiliser une boucle : # X_seq = [] # for i in range(len(X_scaled) - self.lookback): # X_seq.append(X_scaled[i:i + self.lookback]) # X_seq = np.array(X_seq) # # # prédiction # y_pred = self.model.predict(X_seq, verbose=0).flatten() # # # alignement avec les données # preds = [np.nan] * len(dataframe) # start = self.lookback # end = start + len(y_pred) # preds[start:end] = y_pred[:end - start] # # dataframe["lstm_pred"] = preds # # def trainTensorFlow(self, dataframe, metadata, lookback=50, future_steps=1): # # 1) définir la cible # y_values = dataframe[self.indicator_target].values.reshape(-1, 1) # # # 2) définir les features (toutes les colonnes sauf la cible) # feature_columns = self.model_indicators #[col for col in dataframe.columns if col != self.indicator_target] # X_values = dataframe[feature_columns].values # # # 3) normalisation # scaler_X = MinMaxScaler() # X_scaled = scaler_X.fit_transform(X_values) # # scaler_y = MinMaxScaler() # y_scaled = scaler_y.fit_transform(y_values) # # # 4) création des fenêtres glissantes # X = [] # y = [] # for i in range(len(X_scaled) - lookback - future_steps): # X.append(X_scaled[i:i + lookback]) # y.append(y_scaled[i + lookback + future_steps]) # # X = np.array(X) # y = np.array(y) # # # 5) définition du modèle LSTM # model = Sequential([ # LSTM(64, return_sequences=False, input_shape=(lookback, X.shape[2])), # Dense(32, activation="relu"), # Dense(1) # ]) # # model.compile(loss="mse", optimizer="adam") # model.fit(X, y, epochs=20, batch_size=32, verbose=1) # # # 6) sauvegarde # model.save(f"{self.path}/lstm_model.keras") # np.save(f"{self.path}/lstm_scaler_X.npy", scaler_X.data_max_) # np.save(f"{self.path}/lstm_scaler_y.npy", scaler_y.data_max_) # # pour restaurer # # # df = dataframe[self.model_indicators].copy() # # # # # Construction dataset X / y # # X = [] # # y = [] # # # # prices = df[self.indicator_target].values # # # # for i in range(lookback, len(prices) - future_steps): # # X.append(prices[i - lookback:i]) # # y.append(prices[i + future_steps]) # # # # X = np.array(X).reshape(-1, lookback, 1) # # y = np.array(y) # # # # # --- Définition du modèle --- # # model = models.Sequential([ # # layers.Input((lookback, 1)), # # layers.LSTM(64), # # layers.Dense(32, activation="relu"), # # layers.Dense(1) # # ]) # # # # model.compile(optimizer="adam", loss="mse") # # model.summary() # # # # # --- Entraînement --- # # model.fit(X, y, epochs=20, batch_size=32, verbose=1) # # # # # --- Sauvegarde --- # # model.save(f"{self.path}/lstm_model.keras", include_optimizer=False) # # # print("Modèle entraîné et sauvegardé → lstm_model.h5") def kerasGenerateGraphs(self, dataframe): model = self.model self.kerasGenerateGraphModel(model) self.kerasGenerateGraphPredictions(model, dataframe, self.lookback) self.kerasGenerateGraphPoids(model) def kerasGenerateGraphModel(self, model): plot_model( model, to_file=f"{self.path}/lstm_model.png", show_shapes=True, show_layer_names=True ) def kerasGenerateGraphPredictions(self, model, dataframe, lookback): preds = self.tensorFlowGeneratePredictions(dataframe, lookback, model) # plot plt.figure(figsize=(36, 8)) plt.plot(dataframe[self.indicator_target].values, label=self.indicator_target) plt.plot(preds, label="lstm_pred") plt.legend() plt.savefig(f"{self.path}/lstm_predictions.png") plt.close() def kerasGenerateGraphPoids(self, model): for i, layer in enumerate(model.layers): weights = layer.get_weights() # liste de tableaux numpy # Sauvegarde SAFE : tableau d’objets np.save( f"{self.path}/layer_{i}_weights.npy", np.array(weights, dtype=object) ) # Exemple lecture et heatmap weights_layer0 = np.load( f"{self.path}/layer_{i}_weights.npy", allow_pickle=True ) # Choisir un poids 2D W = None for w in weights_layer0: if isinstance(w, np.ndarray) and w.ndim == 2: W = w break if W is None: print(f"Aucune matrice 2D dans layer {i} (rien à afficher).") return plt.figure(figsize=(8, 6)) sns.heatmap(W, cmap="viridis") plt.title(f"Poids 2D du layer {i}") plt.savefig(f"{self.path}/layer{i}_weights.png") plt.close() # ------------------- # Entraînement # ------------------- def trainTensorFlow(self, dataframe, future_steps=1, lookback=50, epochs=40, batch_size=32): X_seq, y_seq = self.tensorFlowPrepareDataFrame(dataframe, future_steps, lookback) # 6) Modèle LSTM self.model = Sequential([ LSTM(64, return_sequences=False, input_shape=(lookback, X_seq.shape[2])), Dense(32, activation="relu"), Dense(1) ]) self.model.compile(loss='mse', optimizer=Adam(learning_rate=1e-4)) self.model.fit(X_seq, y_seq, epochs=epochs, batch_size=batch_size, verbose=1) # 7) Sauvegarde self.model.save(f"{self.path}/lstm_model.keras") # np.save(f"{self.path}/lstm_scaler_X.npy", self.scaler_X.data_max_) # np.save(f"{self.path}/lstm_scaler_y.npy", self.scaler_y.data_max_) def tensorFlowPrepareDataFrame(self, dataframe, future_steps, lookback): target = self.indicator_target # 1) Détecter NaN / Inf et nettoyer feature_columns = self.model_indicators # [col for col in dataframe.columns if col != target] df = dataframe.copy() df.replace([np.inf, -np.inf], np.nan, inplace=True) df.dropna(subset=feature_columns + [target], inplace=True) # 2) Séparer features et cible X_values = df[feature_columns].values y_values = df[target].values.reshape(-1, 1) # 3) Gestion colonnes constantes (éviter division par zéro) for i in range(X_values.shape[1]): if X_values[:, i].max() == X_values[:, i].min(): X_values[:, i] = 0.0 if y_values.max() == y_values.min(): y_values[:] = 0.0 # 4) Normalisation self.scaler_X = MinMaxScaler() X_scaled = self.scaler_X.fit_transform(X_values) if self.y_no_scale: y_scaled = y_values else: self.scaler_y = MinMaxScaler() y_scaled = self.scaler_y.fit_transform(y_values) # 5) Création des fenêtres glissantes X_seq = [] y_seq = [] for i in range(len(X_scaled) - lookback - future_steps): X_seq.append(X_scaled[i:i + lookback]) y_seq.append(y_scaled[i + lookback + future_steps]) X_seq = np.array(X_seq) y_seq = np.array(y_seq) # Vérification finale if np.isnan(X_seq).any() or np.isnan(y_seq).any(): raise ValueError("X_seq ou y_seq contient encore des NaN") if np.isinf(X_seq).any() or np.isinf(y_seq).any(): raise ValueError("X_seq ou y_seq contient encore des Inf") return X_seq, y_seq # ------------------- # Prédiction # ------------------- def predictTensorFlow(self, dataframe, future_steps=1, lookback=50): feature_columns = self.model_indicators #[col for col in dataframe.columns if col != self.indicator_target] # charger le modèle si pas déjà chargé if self.model is None: self.model = load_model(f"{self.path}/lstm_model.keras", compile=False) X_seq, y_seq = self.tensorFlowPrepareDataFrame(dataframe, future_steps, lookback) preds = self.tensorFlowGeneratePredictions(dataframe, lookback, self.model) # # features = toutes les colonnes sauf la cible # feature_columns = self.model_indicators #[col for col in dataframe.columns if col != self.indicator_target] # X_values = dataframe[feature_columns].values # # # normalisation (avec le scaler utilisé à l'entraînement) # X_scaled = self.scaler_X.transform(X_values) # # # créer les séquences glissantes # X_seq = [] # for i in range(len(X_scaled) - self.lookback): # X_seq.append(X_scaled[i:i + self.lookback]) # X_seq = np.array(X_seq) # # # prédictions # y_pred_scaled = self.model.predict(X_seq, verbose=0).flatten() # y_pred = self.scaler_y.inverse_transform(y_pred_scaled.reshape(-1, 1)).flatten() # # # alignement avec les données # preds = [np.nan] * len(dataframe) # start = self.lookback # end = start + len(y_pred) # # preds[start:end] = y_pred[:end - start] # preds[start:start + len(y_pred)] = y_pred # # # # features # # X_values = dataframe[feature_columns].values # # X_scaled = self.scaler_X.transform(X_values) # # # # # création des fenêtres # # X_seq = [] # # for i in range(len(X_scaled) - self.lookback): # # X_seq.append(X_scaled[i:i + self.lookback]) # # X_seq = np.array(X_seq) # # # # # prédiction # # y_pred_scaled = self.model.predict(X_seq, verbose=0).flatten() # # y_pred = self.scaler_y.inverse_transform(y_pred_scaled.reshape(-1, 1)).flatten() # # # # # alignement avec le dataframe # # preds = [np.nan] * len(dataframe) # # start = self.lookback # # end = start + len(y_pred) # # preds[start:end] = y_pred[:end-start] # # # preds[start:start + len(y_pred)] = y_pred dataframe["lstm_pred"] = preds return dataframe def tensorFlowGeneratePredictions(self, dataframe, lookback, model): # features = toutes les colonnes sauf la cible feature_columns = self.model_indicators # [col for col in dataframe.columns if col != self.indicator_target] X_values = dataframe[feature_columns].values # normalisation (avec le scaler utilisé à l'entraînement) X_scaled = self.scaler_X.transform(X_values) # créer les séquences glissantes X_seq = [] for i in range(len(X_scaled) - lookback): X_seq.append(X_scaled[i:i + lookback]) X_seq = np.array(X_seq) # prédictions y_pred_scaled = model.predict(X_seq, verbose=0).flatten() if self.y_no_scale: y_pred = y_pred_scaled else: y_pred = self.scaler_y.inverse_transform(y_pred_scaled.reshape(-1, 1)).flatten() # alignement avec les données preds = [np.nan] * len(dataframe) start = lookback end = start + len(y_pred) # preds[start:end] = y_pred[:end - start] preds[start:start + len(y_pred)] = y_pred return preds